Skip to main content

Full text of "DARPA Technical Accomplishments and Historical Review of Selected DARPA Projects Volume 1"

See other formats

Copy 42 of 70 copies 



Volume I 

Sidney G, Reed 
Richard H. Van Atta 
Seymour J. Deitchman 


Prepared for 
I>efense Advanced Research Projects Agency 

1801 N. Beauiegan) Street, Alexandria, Virginia 22311-1772 

IDA Log No. HQ 89-34081 


IDA pufalUm «» fDllniing doamwnli to nport Hit raaiKt al Mi 

Rtpei ^ aw th« most mttwritotiw and mort ciwfuity camidfd producli IDA puMlihw. 

ynfaody multi ^ maior profKti "rtiich (a) haM a diract taarfng on 
dKWow aftacttnf major procnim, (b) addran isnits of tJgnHtaint cancwn to Ifw 

leant •cenomlc impiicaltons. IDA Reports ara rtvitwvd by oiitsidt phmIi oT wparts to 
•MUto thtlr hifh quaiHy and rrt^^ 
brflwPtoiWmlof UM. 

Group Rgporti 

Group Rapofto i«ofd 1I» liiidlnp «d iwilh 0^ 

'^Ilg!^ °* fcUM di i i li id Jmil Bt major Hmm wWch oMiwwIin would 


Piip< ii,al«B authorltolhwandcafofatlyconsldBf«dp>oductio»IDA.addWM«^^ 
art narrows 

thattti^ moat too W|h rtwKlardi «p«a^ 
or tonnai mgUK/ *~ 



Mifarta nUwawprfcdpooln QuidiroactkMitudiM. (b)torKonltt»pnoaodinparcanl^ 
nmcMMd (e) to mate avallabit pnllminwy and tentotlv* itnrito of anilyin, 

(d) to racart dato dwrafcipod In too couw ol an I n titi M lim. or fa> te i m .!. . ,^^ 111 

««»*^» wtonaliMdand unowluatod. Tha iwiow rtiPA nnriimaim h ■'"■f 
to thair ' ' ■ - 

1!^!!!!!)!^^^^^^ ^^' ^ ^ ^''''''^* ^ ^ 903 84C 0031 

for tho Oopartmont of Dtfanaa. Tha puMl ca tfei i of tola IDA Papor doos not indicate 
M^oraaniam ^ Ow Oapartaiani 
nttteting tht omdai positton of ttiat Aflan^. 

25^5* Jl!!IL!?^ ^ «» teanuto Mial it maoti too itondards of 
condiaiont and ra co ii ii »«i da Uu.t i aw proparty wpportid br tito matortal 

Apprwwd for puWtc 


Form Approvod 
0MB No. 0704^18$ 

ain<pl«tn9 «nd f»w««nfl #»• MlHrtGW o( mtami«an. S«>4 connifn* ngardbg ttM burd« Mma* « «ny o«Mr lapM «f M« HKtidn ol IntomiMM. Mwinfi mmmhm ler f **'t2I!I1 
HMdqMrMra SwwBW. OwdBWIi tar WewMtoi OpwMigw m4 nippm. 1215 <MIWMn 0«m Hi«hii»«tr. S<ii«> IZ04, Min«un. VA 22202-4302. «m« ki «w diet MMO«m«MM« M|« rwin«wli FMKMn ftt^ 


February 1990 


FINAL- -October 1988-December 1989 


DARPA Technical Accomplishments 

An Historical Review of Selected DARPA Projects Volume 1 

C - MDA 903 84 C 0031 
T - DARPA Assignment A-1 1 9 

6. AUTH0R{S) 

Sidney G. Reed. Richard H. Van Atta. Seymour J. Deitchman 


Institute for Defense Analyses 
1801 N. Beauregard Street 
Alexandria. VA 22311 


IDA Paper P*2192 


Defense Advanced Research Projects Agency 
1400 Wilson Blvd. 
Arlington, VA 22209-2308 




Approved for public release; distribution unlimited. 


13. ABSTRACT rAtex//num^wonf«; 

This is the first volume of a planned two-volume history of selected DARPA projects and programs that were 
undertaken from the agency's inception to the present day. The purpose of this history is to record, for projects 
and programs having important outputs and for which adequate and appropriate data could be gathered, the 
chronological and technical histories in such a way that (a) the influence of the projects or programs on defense 
and civilian technology could be traced, and (b) Implementation lessons could be extracted that would help 
DARPA manage future programs in such a way as to enhance their chances of success. 

This volume describes the genesis of the study, the approach that was taken in carrying it out. and program 
histories of 28 DARPA projects. Each history describes the genesis of the project or program, the major 
participants and events in its lifetime, and contains a flow diagram illustrating the complex of interactions among 
organizations over time that characterize the project. Each project review ends with observatioris about the 
project's success and the nature of its impact. Volume II. due In June 1990, will present 27 additional histories, in 
the same format, and will synthesize the obsen/ations about success and influence m such a way that dahpa can 
apply the results to future program management. 


Defense Advanced Research Projects Agency, DARPA projects, 
lessons learned, observations 

15. NUMBER Of KAoea 









indard Form 298 (Rev. 2-69) 

NSN 7540-01 -2«W500 Pr,.«*«ibyANSiaid.z»A 



Volume I 

Sidney G, Reed 
Richard H. Van Ana 
Seymour J. Deitchman 

February 1990 



Coniraci MDA 903 84 C 003 1 
DARPA Assignment A- 1 19 


This is the first volume of a planned two-volume history of selected DARPA 
projects and programs that were undertaken from the agency's inception to the present day. 
The purpose of this history is to record^ for projects and programs having important 
outputs and for which adequate and appropriate data could be gathered, the chronological 
and technical histories in such a way that (a) the influence of the projects or programs on 
defense and civilian technology could be traced, and (b) implementation lessons could be 
extracted that would help DARPA manage future programs in such a way as to enhance 
their chances of success. 

This volume describes the genesis of the study, the approach that was taken in 
carrying it out, and program histories of 28 DARPA projects. Each history describes the 
genesis of the project or program, the major participants and events in its lifetime, and 
contains a flow diagram illustrating the complex of interactions among organizations over 
time that characterize the projccL Each project review ends with observations about the 
project's success and the nature of its impact Volume n, due in June 1990, will present 
27 additional histories, in the same format, and will synthesize the observations about 
success and influence in such a way that DARPA can apply the results to funire program 



Abstract ^ 


Purpose and Scope 1 

Study Approach 3 

Volume n (Proposed Approach and Outline) 9 




Annex ^"^^ 

n. TIROS Weather Satellites 2-1 

ni. TRANSIT Navigation Satellite 3-1 




VI. ESAR Phased Airay Radar 6-1 

VIL TABSTONE Infrared Measurements 7-1 

Vm. High Energy Lasers 8"^ 

IX, OTH Radar ^'^ 

X. AMOS: ARPAMidcourse Optical Station 10-1 

Annex lO-^l 



XL VELA Hotel SateUites, • ....11-r^ 

XIL VELA Uniform: WWNSS ■ 12-1^ 

Xni. VELA Uniform: The Very Large Arrays. LAS A and NORSAR -v:^^'^ 


XIV. Impact on M-16 Rifle t: r'-'-^-^^^^lnll! 

Annex I'm' h 

XV. Camp Sentinel Radar ^ • • ^^-1^ :| 

XVL TheX-26B-QT-2 , ^^^^ f 

XVIL Pocket VETO: Balloon-Bome Radar • 

J. ... 


XVin. ILLIACIV 18"^ 

lo'V , w 

XIX. Project MAC: Computer Time Sharing ; -^^' ^ ^ / .iji^ 

XX. ARPANET r -^^ V^Ji^^^^^ 

XXI. Artificial InteUigence • A , . . .2« pi 

XXIL Morse Code Reader "^^^^ > , 

xxm. ACCAT ..i^-. • ^Sm t 

XXIV. LAMBDA: Large Aperture Towed Aiiays ^ ■! 

XXV. SLCSAT .^^2|>lL V |^ 

■ h 


XXVI. TankBreaker V""?' '^^^ 

XXVn. fflMAG/HSVT-L -v^- ^ S 

XXVHL Mini-RPVs .,...^......,^0 > ^^1, 




DARPA began in 1958 as the Advanced Research Projects Agency (ARPA) with 
the mission of creating a U.S. capabiUty to Uunch and use spacecraft, after the Soviet 
Sputnik launch. Subsequentiy it was given a broader charter, to advance defense 
technology in many critical areas and to help the DoD create military capabilities of a 
character that die MiHtary Sendees and Departments were not able or willing to develop for 
any of several reasons: because the risks could not be accepted within the limits of Service 
R&D and procurement budgets; because tiiose budgets did not allow timely enough 
response to newly appearing needs; because die feasibiUty or military values of the new 
capabilities were not apparent at the beginning, so tiiat the Services declined to invest in 
them; or because die capabiUties did not fall obviously into the mission structure of any one 
Service, so that diere was no eager, ab initio source of support for development and 
operational trials. 

ARPA's charter, therefore, came to include several means by which die agency, 
whose name was changed to the Defense Advanced Research Projects Agency (DARPA) 
in 1972, could undertake new projects and programs. These included assignment by die 
President, the Secretary of Defense or his senior technical subordinates, requests by 
Congress or by the Services, or work undertaken on DARPA initiative (ratified by die 
Secretary of Defense and the Congress in the budget approval process if by no odier 
means) if the agency saw that a miHtary need could be met widi a technological advance 
that was not being explored or exploited. In all die cases related to Service missions, even 
those where Uiere was initial Seivice opposition to an idea, die agency estabUshed some 
appropriate relationship with die Services and MiUtary Departments, as a matter of 
stimulating dieir support, capitalizing on dieir knowledge and often on dieir personnel and 
facilities, and ultimately of interesting diem in using die results of the projects and 
transferring die products to diem for exploitation and use. In odier cases, such as die broad 
DARPA program on nuclear test monitoring, DARPA has established similar relations to 
appropriate non-defense agencies. 


In these modes DARPA undertook, over the years untU now, huiKirc^ ^ 
and programs,! some large and some smaU, in areas such as BalUstic MissUe Defense. ' ;| 
Nuclear Test Monitoring, counterinsurgency warfare in Southeast Asia, advanced ^| 
information processing, advanced naval technologies, advanced technologies appUcable to . 
tactical and strategic land and air warfare systems, and basic research in, such areas as.. ^ ; , 
materials, underwater phenomenology and the phenomenology associatoi'with observation '^-^^^ 
from space, to mention just a partial Ust. The output from these efforts has beeii 
prodigious, and it has had a profound impact on die worid of defense technology and often 
on civilian technology as well. 

One purpose of this task has been to trace that impact It has sought to ton how,a 
representative sampling of projects interacted with the world of "users" to affect thg^ 
technology available to them and how they appUed that technology in systeins.a^: 

In some cases the output of D ARPA projects was acceptet} dirccdy. In others; th^: 
influence of D ARP A projects that were not transferred expUcidy for use may nevothelcssr 
have been felt indirectly in changing the direction of an area of miUtaiy K&D ot die fcnm* V.. 
of military systems as articulated in industry's systems design concfepts and ' 
implementation. In stiU other cases technological advances diat were clear and apparent- 
improvements over earUer approaches emerged from DARPA projects and were adoptpd 
because they did represent such advances. Finally, even some projects diat appear^ ^ 
initially to have been failures have been found on deeper exploration to haye niade 
themselves felt over time in many indirea ways. 

In all cases there were complex interactions among DARPA, the Services, the 
academic world and defense as weU as civilian industry. Given the multifaceted nature of ^ 
die influence DARPA can have in Uie course of these interactions, the tracing of influence 
of DARPA work is not a straightforward task. Views of influence vary widi participating 
individuals, many related efforts outside DARPA interacted widi die DARPA effojs 

UIVUTftUUHM, — , . 

diemsclvcs, and only in some cases is diere a clear padi from genesis Of an idea to its^ 

and apparent use. ^ . 

From this, a second purpose of this task has been to deUneate die n^/i|| 
DARPA s influence and to draw from that lessons diat can help DARPA consciously^, ^ 

A program is a collection of intenelated projects in a single subject area. 


manage the formulation of its program and the guidance of projects so as to maximize the 
acceptance and use of the program's results. Thus, the overall report will describe the 
histories of the individual projects selected for review, and it will draw from die histories 
some lessons that might be learned about success, impact and scientific and technological 
influence. It will then aggregate those lessons into a more complete overview that attempts 
to answer the questions regarding the second objective. 

This is the first of two volumes presenting die histories of specific projects and 
programs, from the point of view of learning how the DARPA efforts influenced the 
defense and civiUan technological worlds. This volume describes 28 projects, grouped to 
correspond to die larger program areas of which tiiey were pan, drawn mainly, but not 
exclusively, from die first two durds or so of die 1958-1988 period. Thus, many of die 
projects reviewed have been completed and die outcome of dieir impact is mosdy apparent. 
Hie projects in dus volume arc Usted in Table L Each project history describes die genesis 
of die project, die major participants and events in its evolution and its appUcations or 
adaptation into odier technical work, to die extent diey are known. Each includes an 
organization/time flow chart diat illustrates die environment and die complex interchanges 
in die project's genesis, execution and influence on odier efforts. Each history ends widi 
brief observations about its "success." 

Volume n, to be published in June 1990, wiU present 27 additional project 
histories, listed in Table H. in die same format and will also include brief accounts of die 
broader programs' histories, and a comprehensive analysis of die lessons about die extent 
and success of technology transfer, and die influence of DARPA, diat have been learned * 
fiom reviewing die histories of all die individual projects. 


The projects studied were selected by die IDA project team and DARPA 
management working togedier. based on two criteria: (a) dieir importance, judged on die 
basis of evidence in attestation and documentation; and (b) die expected availabiUty of data. 
The data available would have to contain sufficient information to permit elucidation of 
D ARPA's role and contribution, tracing die padis of technical events dirough ultimate use, 
assessment of die impact and spin-offs of die output, and clear enough records to permit 
evaluation of lessons learned from die outcome. The Usts shown in Tables I and H resulted 


Table 1. DARPA Projects Reviewed In Volume I 


































Table 2. DARPA Projects to be Reviewed In Volume II 



























from several iterations to ensure that the selection cxiteria, especially the second, could be 


The starting point was a Ust of accompUshments that D ARPA had prepared for the 
Agency's 25th anniversary celebration. Most topics on this Ust are single projects, but 
some arc groups of projects, constimting sub-programs under a broader program area 
(such as DEFENDER). This Ust, which had inputs from former DARPA Directors and 
current and former program managers, formed the woridng basis for discussions between 
the IDA project leader and DARPA management DARPA was amenable to changes that 
either added to or subtracted from the list, depending on what preliminary explorations 
showed about data availability and revised perspectives on the value of the programs- 
outputs, m resulting list was then divided into those entries that could easUy be described 
from data that were mainly available, and others for which extensive research would be 
necessary to cUcit the facmal histories. Botii kinds of descriptions are included in this 
volume; the division simply meant that some of the project reviews on the agreed list had to 
be postponed until Ac next volume of tfiis report could be completed. 

The facmal histories of the selected projects or programs were cUcited from a 
combination of sources: interviews with participants, reference to DARPA records, review 
of tiie technical Uteraturc, congressional hearings, and interviews with other individuals 
who had first-hand knowledge about at least some aspects of the projects. Aiftertiie 
relevant facts and judgments were obtained from these various sources, the flow charts and 
the histories were prepared. 

Available data included a Ust. prepared by Mr. A. Van Avery, a former ..ARPA 
program manager, of ARPA or DARPA Orders up to 1975,2 and a compUation by 
DARPA of the acmal ARPA or DARPA orders tiiat had been issued from 1975 through 
1988. There were also compilations by the BatteUe Memorial Instimte of one-page project 
descriptions for the projects in the DEFENDER and AGILE programs, prepared under 
DARPA contract BatteUe had also prepared a categorization and Usting of all the DARPA 
programs for several years in the mid 1970s. Other documentary sources included Service 
program histories, a book about die VELA program;^ a book by Dr. H. Yoric. the Chief 

ARPA or DARPA Ordeni are documents signed by the Director that convey agency funds to contracting 

agencies of the Services who support DARPA administrauvely. 

A. Keir. ed.. The VELA Program, Defense Advanced Research Projects Agency, 1985. 

Scientist of ARPA at its inception;^ a history of ARPA up to 1974;^ Congressional 
hearings for the relevant years; and access to the DARPA and IDA archives. 

Interviews with participants in or observers of the projeas or programs being 
described, and of follow-on or related Service or commercial impacts, were undertaken 
wherever the documentary record was not clear and complete. The interviews were used to 
gain insights and clues as to where to seek further data, but the ensuing written descriptions 
of die projects were based to the greatest extent possible on die written record. The 
interviews furnished valuable information for corroboration or illumination of documentary 
data, and in such cases the resulting interviews were used and appropriately footnoted (as 
were the documentary sources). Interviews were often most useful in gaining insights on 
die subsequent impact or transfer of DARPA technology in both die military and 
commercial arena. Therefore, we make a expUcit effort to obtain the perceptions of those 
outside of DARPA who were knowledgeable about die program, its origins and related 
research supported by others. 

DARPA history and D ARPA-relatcd individuals were not die only sources for die 
descriptions, since ARPA or DARPA influence on events and systems elsewhere in die 
DoD and commercial worlds was also being sought rnttuence works in two directions, 
including diat exerted upon DARPA as well as diat exerted by DARPA, and appropriate 
data from outside sources were gadicred and used in die same manner as die DARPA or 
DARPA-rclated data. A good example is die description of die development of Over-The- 
Horizon radar, where die Australians have written dicir own history of dieir work in dus 
area and participation in the joint U.S.-Australian program. 

An attempt was made to estimate die costs of die ARPA or DARPA projects for 
comparison widi dollar figures relating to dieir impacts. Congressional hearings and 
D ARPA records were die information sources for costs. This information was used where 
it was readily available and appeared credible. 

While we believe diat die accounts resulting from die process described are as 
accurate as die overall project-based approach, available time and information permit, 
experience has shown diat new insights and information are discovered continually on 
diesc topics, at unpredictable times after work on diem begins. Often die unearthing of 
information on die evolution and subsequent effects of a project is akin to sleuddng or 

^ Herbert F. York, Making Weapons, Talking Peace, Basic Books, New York, 1987. 

5 The Advanced Research Projects Agency, 1958-1974, Richard A. Barber Associates. 1975. 


prospecting with leads playing out or becoming blind alleys. Frequendy the sources of 
information are obscure conference papers or documents diat may take several weeks to 
obtain. Moreover, more than once important information on die impact of a DARPA 
project was gleaned ftom documents being reviewed for assessing anoUicr DARPA project 
Additionally, the more recent efforts have not yet fuUy run dieir course. Thus. Volume H 
may contain additional information that appears after pubUcation. about the project and 
program discussions in tiiis volume, and the sponsor may wish to update the entire report 
every few years as the outputs of die program are used more and insights about their 
importance change. 

Every attempt has been made to keep die project or sub-program discussions 
unclassified. While omission of classified information necessarily makes die account of 
events incomplete, it was believed diat technical detail, which tends to constitute die 
classified component of a project, was less important dian scientific and engineering 
principle and die simple flow of ideas, events and technical interactions among different 
programs and groups; die latter set of concems sh^cd die main avenues of investigation. 

The results of die effort to date arc given in die program assessments of Volume I 
for die 28 projects Usted in Table I. in die order and in die program groupings shown in 

table. The Ust is organized by program categories, widi projects listed under dicm, in 
rough historical order. Some of die projects and sub-programs to be described in Volume 
n will predate some of diose in dns volume, and die order will be rearranged as appropriate 
for the final history. 

It should be noted diat dus volume, and die one to follow, do not constitute 
histories in die true sense of die word, nor do diey, togedier, constitute a complete and 
balanced history of die agency. Moreover, while we have grouped die projects und«r die 
broad program headings to which diey mainly belonged, it is important to note diat a 
description of some of die projects in a broad DARPA program area may not convey an 
adequate sense of die overall strategy and impact of die programs. However, die indi^ddual 
narratives describe a selected set of projects and programs chosen because it was beUeved 
diat diey were important in die relationship of die agency widi die development of technical 
capabilities in die "outside world," and because it was believed diat dieir importance could 
be traced and documented. Many important gaps remain to be filled-for example, die 
materials area, some major aspects of die DEFENDER program, and odiers. Many of 
diese will be fiUed by die added project and program descriptions planned for Volume H. 


Thus, wc do not represent this document as a definitive account of all ARPA and 
DARPA activities or of the overall impact of the broader programs since the agency's 
inception. But we believe it constitutes a useful working document that the sponsor can 
apply to current and planned activities and update as new information arrives. 

We have made a special attempt, in the time available, to have Volume I reviewed 
by knowledgeable individuals who could judge its accuracy overall or in part The entire 
document was reviewed by R. Sproull, C. Herzfeld. E, Rcchtin, G. Heilmeier, 
S. Lukasik, and R. Cooper, all ex- ARPA or DARPA directors, and also by F. Koctiicr and 
A. Flax. Parts of Volume I were reviewed by H. York, C.W. Cook, MGen. J. Toomay, 
T. Bartee, R. Finkler, J. Kreis, Capt. H. Cox, O.G. Villard, T. Croft, R. Schindler and 
H. Wolfhaid and R. Collins. We thank the reviewers for their comments and insights, 
which have greatiy benefited the document Any persistent errors remain the responsibility 
of die authors. 

VOLUME n - Proposed Approach and Outline 

Based on die work done to date, wc have developed some preliminary ideas for 
assessing die overall impact of die identified DARPA projects. Our major concern is that 
any such assessment appreciate (1) the complexity of the research undertaken by DARPA 
and (2) the range of potential impact this research might have. Our experience on this 
subject is that individuals, within DoD as weU as elsewhere in government and industry, 
frequendy define DARPA's role very expUcidy and narrowly and define "success" based 
on such mterprctations of DARPA's role. Given die history and charter of DARPA, die 
multifaceted nature of die work diat it has been assigned as well as initiated itself, such 
narrow concepts are not apt Sometimes diey lead to misplaced criticism or self-flageUation 
for programs not direcdy leading to a fielded weapon system We contend diat technology 
transfer, while an important, issue and an important basis for judging DARPA's 
accompUshments, must not be conceived too narrowly. On die odier hand, it is inherent to 
sound management principles, even in an advanced research enterprise, to demand Uiat 
programs be conceived, overseen, and ultimately judged on die degree to which diey will 
make a difference to die accomplishment of die overall organization's objectives and 
missions. It is in dus sense diat we will review and assess die accomplishments of 



A. ASSESSMENT OF SUCCESS - Will assess and aggregate across DARPA 
programs to determine factors that differentiate degree and type of success based 
upon the following: 

1 Origin of Program - How did it get to DARPA and did its origins have any 
impUcarions for success? e.g., "Project was White House initiative of highest 
priority," or "Project was brought to DARPA by Service research office after 
failing to get funding ficom Service." 

2 Objective of Project - What was the initial objective? Was it to develop a 
miUtary system? Assess the potential of a new technology for improving a 
military capabiUty? Was it aimed at improving a technology base for potenual 
defense application? 

Did it stay die same? If it changed, why? Was objective clear, specific? Was 
it broad, general? 

3. TypeofProgram 

. Mission or Operational Program (type: Nuclear Detection, Space Payload, 

• Weapons Research and Development (Strategic & Tactical) 

• Information Systems R&D (type: C^I, etc.) 

• Technology Base stimulation/exploration (assess new technology to guard 
against surprise and identify potential, push technology application for 
defense use, overcome obstacles to technology development) 

4 . Status of Technology 

• U.S. leadership position relative to adversaries 

• U,S. falling behind or trailing relative to otiiers 

5 . Political-Organizational Gimate/Environment 

• Defense Transfer - Competing with other approaches or appUcations of 
user versus cooperating with or supported by user 

. External factors - Create resistance versus facilitate 


6. Type of Success - The results and impact of the DARPA programs will be 
characterized according to the following categories (these are not mutually 
exclusive and are subject to revision): 

• DARPA-developed system itself actually fielded for military, 
defense, or national security mission. Still used? If not, why? 
Obsolete and replaced. Threat changed Superseded by another 
technology (DARPA role?) 

• DARPA-developed system transferred to military service or 
agency and fielded, etc. 

• DARPA-developed system concept transfeired to Service (or 
Agency) for further development and subsequent fielding 

• DARPA-developed system concept transferred to Service (or 
Agency) for further development (and subsequent fielding?) 

• DARPA-developed technology used, adapted, by Service or 
Agency in development of weapon system or defense application 
(subsequendy fielded?) 

• DARPA development achieved quanmm jump in fundamental 
scientific or technical knowledge of use to defense or broader 

• DARPA research stimulated or explored nascent, high potential 
(or unknown potential) technology and related technology base to 
determine military worth and/or degree of adversarial thixsat 

• DARPA sped up the development of a technology (by several 
years) for meeting defense application 

• DARPA research led to substantial spin-offs/spill-overs to other 
military systems, commercial applications, and/or overall 
technology base significant for defense or national security 

• DARPA research caused fundamental rethinking, redefinition of 
defense mission or approach to a mission (with major impact on 
alternative systems) ^ 

• DARPA research had widespread indirect, but identifiable 
payoffs, e.g., pervasive impact on technology area; established 
new technology base which has led to many, perhaps unforeseen, 
improvements in national defense and econonnic capabilities 



WiU summarize aspects of DARPA's successful accomplishments that can be useful 
for selecting and conducting programs in the future. Can such "successes" be 
repeated in today s environment? Are there differences in types of programs that 
lead to differing kinds and degrees of success? Are there indications of precursors, 
minimum requirements, ideal conditions for success? Given DARPA's tnission 
(high risk-high potential), how assured should success be? (Does analysis show 
examples of "success" that were aimed too low?) 


For Volume n. 27 DARPA projects wiU be reviewed and organized as Usted in 
Table n above. 




1... -j^'ti iCL 



The ARGUS experiment was one of the earliest major ARPA space projects, 
involving nuclear explosions at altitudes in the hundreds of kilometers, with a coordinated 
set of measurements by sateUites, rockets and ground stations. It was a test of die concept 
that large numbers of electrons might be injected into the eanh's magnetic fields, be tr^ped 
there, and affect baUistic missile warheads, satclUtes, and jamming of radio and radar 
systems. The experiment was accompUshcd in six months in response to a Presidential 
order. ARGUS was a very risky,very large scale, and quite successful project, getting 
ARPA off to a good start. 


The ARGUS concept was suggested by the late Nicholas C. Christofilos, then at 
AECs Uvermore Laboratory, in reaction to the advantage in space the Soviets had shown 
by their launch of the Sputniks in late 1957. At Uvermore, Christofilos was involved in 
the ASTRON project to trap and heat hydrogen ions in a magnetic field formed by a 
toroidal current of electrons, for controlled thennonuclear fusion. According to a recent 
account,! Christofilos* suggestion was: 

an Astrodome-like defensive shield made up of high-energy electrons 
trapped in die earth's magnetic field .... in essence, he proposed to explode 
a large number of nuclear weapons, thousands per year, m the lower part 
of the earth's magnetosphere, just above the upper reaches of the 
atmosphere. These explosions would produce huge quantities of reactive 
atoms and these in turn would emit high-energy electrons (beta particles) 
and inject them into a region of space where the earth's magnettc field 
would trap and hold on to tiiem for a long time ... months or longer. 

The number of trapped electrons, he believed, would be enough to cause 
severe radiation damage-and even heat damage-to anything, man or 
nuclear weapon, tiiat tried to fly through the region. He expected that this 
region would extend over the whole planet, save only a relanvely smaU 
region, around each pole. Nick had, in effect, invented a version of the 

1 H. Yoik, in Making Weapons, Talking Peace, Basic Books. New York, 1987, p. 130. 


neutraUy occumng Van Allen belt, before it was discovered. He proposed 
an exp^ment. nlned Argus, ... in it we would explode a nudear bomb 
high above the atmosphere, after first placing in orbit a sattUite wift 
SlfmS on board suitable for observing the predicted injecnon of high 
energy electrons in the magnetosphere. 

Christofilos- idea was brought to the attention of the then recently formed 
President's Science Advisory Committee (PSAQ by Dr. H. York, then director of the 
Livermore Laboratory and a member of PSAC According to James Killian, Jr., then the 
President's Science Advisor* 

PSAC strongly supported a test of this theory. It felt that the testjTO»}d 
vi^ld ii^pormm ne^entific knowledge about the earth> "M^^P.^' ^ 
Ke ^vior of radiation in space, the test might provide <^tt and hdp 
answer questions that were under debate. Would such an W^afn 
electroi^ interfere with radar and radio, m^ght *c tnan-mdj^ced cu^ 
suggest any possibilities for an anaballistic missile system? What wouW be 
the effects of such an explosion on our early-warning and global 
communications systems? Qeariy there might be 
achieved by such a test.. JSAC recommended thatthe great eJ^en^ent be 
undertaken. Apparently for security purposes the President Pfef^Tf 
have the matte? discussed at an NSC meenng. I Presenwd.*' .fSAC 
recoinmendarion to him on 1 May 1958 and he made the decision himself 

that the experiment be undertaken. 

At the time Christofilos presented his ideas and proposals* it was not at aU clear 

how these could be canied out York^ says: 

The experiment he wanted was on a grand scale and necessarily involved 
sateUitos. Such devices were coming along, but we had not yet flown an^ 
Argus, to say the least, was a collection of far out mterestmg jdeas but it 
seemed there was simply no place to take an invenoon l^f Nick^ Be^^^ 
such an invention and the experiments that supponed it could be ac^^d w 
a whoUy new organization had to be created, o?<^^,tf°"^tn S^^^^ 
projects of this grand scope and great novelty, projects that had to be taken 
seriously but did not fit into any existing niche. 

ARPA was this new organization, and York became its Chief Scientist in March 

1958. Once there he had: 

both the responsibility and authority for carrying out die c^Pf^n^^^^^^ 
Christofilos and I had first discussed four inontiis earher. With the help of 
Nick himself, we were able to elaborate ARPA Order #4,^ conveymg fiscal 

2 James R. Killian. Jr. "Spuuiik. Scicntisis and Eiscnhowcn" MIT Press 1977» p. 

3 York, ibid., p. 131. 

4 Dated 4/28/58. 


authority and instructions to the Anned Forces Special Weapons Project, 
and thus to set in motion Project Argus. 


Regarding the scale and plans for the project Killian says: 

Obviously the test would require immense resources and facilities involving 
both the Atomic Energy Commission and the Department of Defense and a 
group of other organizations. As finally organized, the opcranonal and 
technological management of the project was vested m the Advanced 
Research Projects Agency (ARPA) of the DoD. The nuclear explosions 
would be provided by the AEC, the Explorer rocket by the Army Center m 
HuntsviUe, and the Navy would provide tiie task force. The Air Force 
Special Weapons Center undertook the preparation of a senes of high- 
altitude sounding rockets for die smdy of die lower fringes of the expected 
cffects-at altiwdes of about 500 mUes using a five-stage sohd-propellant 
rocket vehicle that had been developed by the NACA. The Air Force 
Cambridge Research Center and the Stanford Research Institute developed, 
located, and prepared to operate a variety of equipment at smtablc ground 
stations and aboard aircraft and ships. In his capacity as Chief Sciennst of 
ARPA, Dr. York directed the program and provided a link with the Science 
Advisory Committee. The Navy was entrusted with the execution of the 
experiment ... three rockets were launched from the roUing, pitching base 
of the Norton Sound and all these were successful in deUvcnng the nuclear 
test devices. 

The detailed organization was handled efficientiy by an informal group consisting 
of Dr; Frank Shelton, Chief Scientist of the Armed Forces Special Weapons Project 
(AFSWP), and CoL Dent Uy of ARPA (ex-deputy chief of AFSWP). Since AFSWP was 
occupied in the conduct of the TEAK and ORANGE tests (megaton level and high altitudes 
< 100 km) in die Pacific in July, a new ARGUS task force was formed by the Navy, and it 
rendezvoused with the U.S.S. Norton Sound (which had sailed from the Pacific Coast) in 
die Soudi Adantic on August 25.^ 

About what happened, York says': 

Between August 27 and September 6, 1958 diree nuclear weapons were 
exploded above die atmosphere at an altimde of diree hundred nules alx)ve 
die Soudi Adantic at a point approximately longimde ten degrees west and 
latimde forty degrees soudi. A sateUite, Explorer 4, suitable for observing 
die high energy electrons produced by die explosion and trapped by Uie 
earth's field, was in place... The bombs had been lofted by a rocket 

5 Killian. ibid., p. 188. 

6 Testing Moratorium Years 1958-61," unpublished manuscript by Dr. F. Shelton. Discussion widi 
Dr. Shelton 7/88. 

7 York, ibid., p. 149. 


launched from a ship in the lee of Gough Island,^ an uninhabited Bnush 
possession located in just the right place in the South Adanuc, for reasons 
having to do with the imperfect symmetry of the earth s (magnetic) ticia. 

More scientific detail, as weU as an interesting account of the scientific background 
at this time and of his own personal involvement, has been published recently by Dr. James 
Van Allen.9 From data gathered earUer from their Explorer I and m satellites, Van Allen 
and his group had concluded that there was trapped radiation in the magnetic field of the 
earth giving a radiation intensity at least a thousand times greater than the cosmic radiation,, 
in what is now known as the "Van AUen belt". Figures 1 and 2, from Van Allcn,io give 2- 
and 3-dimensional pictures of the Van Allen bell regions. Van Allen states: 

In mid April 1958 I informed Pickering and Panofsky of my^^y j*^^" 
reasonably firm interpretation of the observations by Explorers I and m, 
namely that there was a huge population of electrically charged pamcles 
already present in trapped, Stormerian orbits in the earth's external magnetic 
field In the context of our earUer studies of the primary auroral radiauon, 1 
considered it likely that these particles had a natural origin. 

Some of those who knew of Christofilos' ideas suggested, at the time, that this 
trapped radiation might have been due to insertion of electrons by earUer nuclear explosions 
conducted by the Soviet Union.^i 

For the ARGUS experiments, Van Allen's group designed and constructed 
Explorers IV and V. These Explorer saieUites were also sponsored by the International 
Geophysical year QGY), Explorer IV OGY-designated 1958e) was launched in July 1958, 
by an Army Jupiter C. Explorer V did not achieve ori)it Van Allen^^ also makes it 
clearer why the Navy was so involved, and in the South Atlantic: 

From a geomagnetic point of view the best site for the injection of electrons 
into durable orbits was near the geomagnetic equator in the South Atlanac. 

8 In the scientific account of the experiment only the fust of the three explosions is given as occurring in 
ihc vicinity of Gough Island (12« W. SS^ S). The second and third locations are said to have been 
8'' W 50* and 10° W. 50* S. The lec of Gough Island was also used tt> avoid large ships mooons in 
the heavy seas. The two other locations were selected to separate the artificial electron belts and 
improve the measurements at the conjugate points near the Azores. 

9 James A. Van Allen. Origins of Magnetospheric Physics. Smithsonian Institution Press. Washington. 

1983, Chapter Vffl. 

10 "The Argus Test," Van Allen, ibid-, p. 66. 

1 1 Van Allen states that the Soviet scientists had the same idea about the U.S.. ibid., p. 83. 

12 Van AUen, ibid., p. 74. 


Figure 1. A meridian cross-section of contours of equal Intensity of 
geomagnetlcally trapped radiation based on data from Explorers I, Hi, and IV and 
Pioneer Hi. The semicircle at the left represents the earth, and the two undulating curves 
that traverse the diagram represent the outbound (upper curve) and inbound (lower curve) 
trajectories of Pioneer III, The labels on the contours are counts per second of a heavy 
shielded miniature Geiger-Mueller tube. The linear scale of the diagram is in units of the 
earth's equatorial radius (6,378 km). The two distinct regions of high intensity (cross- 
hatched) are the inner and outer radiation belts, separated by a region of lesser intensity 
called the slot. From Van Allen, ibid. 

Figure 2. An artist's tliree-dimenslonai conception of the earth and the Inner 
and outer radiation belts. From Van Allen, ibid. 


Because of the eccentricity of the earth's magnetic field, a site at that 
longitude could minimize the altimde at which injection had to occur^tiile 
an equatorial site could maximize the efficiency for injection m order to 
produce durably trapped orbits. Launching from a ship m an isolated sue 
was desirable because it aUowed the secrecy of the operation to be sale- 
guarded. Two satclUte launchers and three bomb injections weiejudged to 
be the minimum effort to give reasonable assurance of success. TTie Navy s 
guided missile ship, the U.S.S. Norton Sound, which we had mitaated 
with Aerobec rocket launchers in 1949. was selected to launch the rockets. 

The Norton Sound had been used in previous rocket launchings and had an on- 
board computer system to control launch at minimum pitch and roll conditions. 

Important information for closer deiennination of the desirable test location was 
generated from Explorer IV.i^ 

On the basis of the fu^t few weeks of data from Explorer IV, we had 
advised ARPA of a discovery of a minimum in the previously present 
radiation when intensity was plotted against latitude. This finding was 
utilized in helping select die latimde for the ARGUS bursts so that the 
artificial radiation belts would enjoy the optimum prospects of detecaon. 
This choice of latimde nimed out to be the best possible choice withm the 
latimde range of Explorer IV.. i.e., in the "slot" between the previously 
observed 'inner' radiation belt and the newly discovered outer radianon 

Besides the general atmosphere of urgency and desire to "catch up" with the 
Russians, there were xdok definite time constraints. Killian says:^ "The whole program 
was under great pressure to meet deadlines, particularly the deadline for the voluntary one- 
year cessation of nuclear tests that die United States had committed itself to as of 
Oct 31. 1958." 

The problems of such a tight schedule and remote location of launch desired for the 
experiment did not seem at aU attractive to Uiose in the Air Force and Army associated with 
the major rocket development projects at the time. Despite the difficulties of launch at sea, 
and with a strong desire to become involved, the Navy took on the launch task. Dr. Willis 
Hawkins, of Lockheed, has described the rockets used on the Norton Sound, which were 
modifications of die Lockheed X-l? used in previous reentry body experiments, in an 

13 Van Allen, ibid., p. 78. 

14 Killian, ibid., p. 189. 


interesting account of ARGUS which gives the flavor of some of the risks involved,!^ 
Three X-17*s were put on the Norton Sound for the launches, in the hope that at least one 
would be successful. Under way, however, three different altimdes were ordered for the 
explosions. To comply, each X-17 had to be launched successfully at a different angle; 
remaikably, each was successful. 

Van Allen also compares the other nuclear tests in the Pacific shortly before 
ARGUS, with the altitudes of the ARGUS explosions.^^ 

The AEC/DoD tests group successfully produced two bursts of (in the 
megaton yield range) bombs, called Teak and Orange, on August 1 and 
August 12 at approximate altitude of 75 and 45 km, respectively, above 
Johnston Atoll in the Central Pacific: The three Argus bursts (1-2 kiloton 
yield range) were produced successfully on August 27, August 30 and Sept 
6 at altimdes of about 200, 250, and 480 km.^^ 

The Air Force Weapons Center rocket measurements at Wallops Island, Puerto 
Rico and C^pe Canaveral were also able to determine the difference in injection altitudes 
very shordy after the explosions from their measurements and theoretical woik.1^ 

Regarding the outcome: KiUian says^^: "Staggering in scale and complexity, it was 
a beautifully managed and highly successful experiment from beginning to end." York 
says:20 'Ten months from the germ of an idea to its acmal execution in outer space was 
nothing short of fantastic even then; today, with more complex rules and regulations, it 
would be utterly impossible." However, mainly because of the time schedule, scientific 
instrumentation involved was quite limited.^^ 

A comprehensive review of all the ARGUS results took place at Livermore in 
February 1959. The New York Times "broke" the previously classified story in March 

15 Willis Hawkins; Annex to this chapter. Another detailed and flavorful account of the Air Force's 
Weapons Laboratories ARGUS lOCket project, which was conducted with NASA assistance, is given in 
"A New Dimension-Wallops Island Test Range, the First 15 Years." by J.A. Shortal. NASA 
Reference PubUcaiion 1028. 1978 pp. 573-580. 

1« Van Allen, ibid,, p. 78. 

1 Hawkins. Appendbt A. however, indicates that some of these ahitudes may be in question. 
IS Discusskm with Dr. Lew Allen. 8/88. 

19 Killian. ibid., p. 189. 

20 York, ibid., p. 149 

2 1 Some later critics staled ihat ARGUS was pooriy instrumented. Cf. "United States High Altitude Test 
Experiments." Los Alamos Report LA 6405, by H. Hoeriin. Oct 1976, p. 46. 


1959, and an unclassified seminar was held at the National Academy of Sciences at the end 
of April 1959.22 The public statement by the Academy said 23 

A fascinating sequence of observations was obtained. The brilliant initial 
Hash of STe bi^st was succeeded by a fainter but persistent auroral 
luminescence in the atmosphere extending upwards and downwards along 
S^^etic line of force tLugh the burst pom^ '^^tl^fn^^rh^ 
the point where this line of force returns to the earth's atmosphere m 
northern hemisphere-titc so-called conjugate pomt-ncar the Azores Island, 
a bright auroral glow appeared in the sky and was observed ^^^^ 
previously stationed there in anticipation of the event, and the complex 
^ries of recordings began. For the first time m history measure 
geophysical phenomena on a world-wide scale were being related to a 
quantitatively known cause-namely, the injection mto the earth s magncac 
field of a known quantity of electrons of known energies at a known 
position and at a known time. 

The diverse radiation instruments in Explorer IV ^corded and reported to 
ground stations the absolute intensity and position of this sheU of high 
fnergy electrons on its passes through the sheU shortly after the bupts Jl^e 
sateUite continued to lace back and forth through the man-made sheU of 
trapped radiation hour after hour and day after day. Tht physical shape and 
poskion of the sheU were accurately plotted out and the decay of intensity 
was observed. Moreover, the angular distribution of the radianon was 
measured at each point The shape and form of a selected ^apietic shell ot 
the earth's magnetic field were being plotted out for the f^^^^ 
experimental means. In tiieir helical excursions within this shell the flawed 
cl^trons were travcUng vast distances and were foUowmg the magncuc 
field pattern out to altitude of over 40,000 miles. 

York says, briefly. "We found that electrons were trapped as Nick had predicted 
but that they did not penist for as long as he had hoped."2* 

Van Allen gives more scientific detail and outlines the impact on magnetospheric 


22 Quoted by Killian. ibid., p. 190. Piocecdings of .his symposium were Publish^ in Oie °f 
tC^Mational Academy and in the Journal of Geophysical Research, Vol, 64, 659. pp. 869-937. 
SLSn^f ^^St^-l^^^ in Heating of the House of R^P«^"Jf ^^^^^^^^ °" 
™e and Astronautics, 10 April 1959. and in "A Scienust at the White House, by G.B. 
Kistiakowsky. Harvard U. Press, 1976, p. 72. 

23 Quoted by Killian, ibid., p. 190. 

24 York, ibid., p. 149. 

25 Van Allen, ibid., p. 78. 


Figura 3. The narrow dotd>l* spikes in the responses ""^J'""", 
delators were observed af about 0510 GMT on August 30. f»P'<»'*' 
IV traversed the shells of energetic electrons Injected '"«« ♦'•PP«<* .^J"' 
the Argus I burst on August 27 and the ARGUS ii burst about three 
hours earlier on August 30 (From Van Allen, Ibid., p. 79). 


We observed with Explorer IV the effects of all five of the bursts in 
populating the geomagnetic field with energetic electrons. Desjnte the large 
yields of Teak and Orange, the incremental effects on the existing 
population of trapped particles were small and of only a few days lifettme 
because of the atmospheric absorption corresponding to the low alUtudes oi 

The three higher-altitude ARGUS bursts produced clear and weU-observed 
effects (see our Fig. 3) and gave a great impems to understanding 
geomagnetic trapping. About 3% of the available electrons wctc mjected 
\kto durably trapped orbits. The apparent mean lifetime of the first two.ot 
these artificial radiation belts was about three weeks and of the third, about a 
month. In aU three cases a well-defined Stormenan shell of amficiaUy 
injected electrons was produced. Worldwide study of these shells provided 
a result of basic importance-a full geometrical description of the locus ot 
trapping by "labeled" particles. Also, we found that the physical nature of 
the ARGUS radiation, as characterized by our four Explorer iv detectors, 
was quite different than that of the pre-ARGUS radiation, thus dispelling 
the suspicion that the radiation observed by Explorers I and Ul haa 
originated from Soviet nuclear bomb bursts* 

During the approximate month of clear presence of the three artificial 
radiation belts, there was no discernible radial diffusion of the trapped 
electrons, thus permitting determination of an upper Umit on the radial 
diffusion coefficient for such electrons. The gradual decay m mtensity was 
approximately explicable in terms of pitch angle scattering in the tenuous 
atmosphere and consequent loss into the lower atmosphere. 

A comprehensive ten-day workshop on interpretation of die ARGUS 
observations was conducted at Uvcnnore in February 1959. The Physical 
principles of geomagnetic trapping were greaUy clanfied at this woricshop. 
To us, one of the principal puzzles had been the durable integnty of a thin 
radial shell of electrons despite the irregular nature of the real geomagncnc 
field and the existence of both radial and longitudinal drift forces resulting 
from gradients in die magnetic field intensity. We had previously 
understood the importance of the first adiabatic invanant of Alfvcn m 
governing trapping along a given magnetic line of force and the effects to 
die radial component of the gradient of the magnetic field intensity B in 
causing longimdinal drift in an axially symmetric field. But the longimdmal 
component of the gradient of B seemed to imply irregular dnft in radial 
distance and hence in radial spreading, contrary to observanon. The puzzle 
was immediately solved by Northrop and Teller who mvoked the second 
and third adiabatic invariants of cyclic motion to account for me 
observations. These theorems had been proven previously by Rosenbluth 
and Longmire [1957] and applied to plasma confined by a laboratory 
magnetic field. A specific application of these principles was Mcflwam s 
[1961] concept of the L-shell parameter for the reduction of three- 
dimensional particle distributions to two-dimensional ones-a concept mat 
has peraieated the entire subsequent literature of naagnetosphenc physics. 

The adiabatic conservation and nonadiabatic violation of these three 
invariants have proved to be central to understanding trapped pamcle moaon 


and to play a basic role in all of magnetospheric physics. In effect, they 
supplant the rigorous integral of motion found by Stormer for an 
axisymmctric magnetic field and make it possible to understand trapped 
particle motion and the diffusion of particles when the conditions for 
conservation of the three invariants are violated by time-varying magneuc 
and electric fields. The three invariants correspond to the three forms of 
cyclic motion, witii quite different periods, into which the Stormerian 
motion of a charge particle in an approximate dipolar magnetic field can be 
analyzed. The first is the gyro motion of the particle around a field line; the 
second is the latimdinal oscillation of the guiding center (the center of the 
cylinder on which tiie helical motion of the particle occurs) of the particle's 
gryo motion; and die tiiird is the time-avcraged cyclic drift of die guidmg 
center through 360** of longitude. 

The Kirtland rocket measurements were generally consistent with our 
Explorer IV measurements but added important detail on particle 
identification and energy spectra. Also, atmospheric luminescence of 
auroral character was observed along the lines of force on which die bursts 
occurred; an artificial auroral display was observed at the nonhem 
geomagnetic conjugate point of die third burst; radar reflections from die 
auroral mbes of forces were observed in all diree cases; and a vanety of 
transient ionospheric effects were detected. No electromagnetic (cyclotron) 
emission fix)m the trapped electrons was observed by ground stations, a 
result consistent with estimates of the intensity relative to cosmic 

The Livermore meeting recommended furdier research, particularly on methods of 
achieving higher efficiency of injection of electrons into trapped belts. This led to plans for 
a foUow-on test, WILLOW, and some further laboratory and theoretical woric^e but dus 
area was not pursued intensively after the test moratorium in 1958. There were no further 
nuclear explosions between 1958 and 1961. However, four high altitude nuclear 
explosions, one U.S. and dirce Soviet, occurred in 1962. The U.S. "STARHSH" event, 
a 1.4 megaton detonation at an altitude of 400 km near Johnston Island, led to an intense 
artificial radiation belt widi the longest "mean lifetime," ncariy 1.5 years.^^ The intensity 
and lifetime of diis "STARHSH" belt seems to have been somewhat unexpected.^* This 
effea has been partiaUy attributed to magnetohydrodynamic migration outward of die bomb 

26 R Shelton, ibid, and AO 6 Tasks 37^1 of 5/59. 

27 "Spaual distributions and time delay of the intensities of geomagnetically trapped electrons from th^ 
high altitude nuclear burst of July 1962," J.A. Van Allen, in "Radiation Trapped in The Earths 
Ntognetic Field," B.M. McConnac. Ed.. ReideU 1966. p. 577. TTie decay is apparentiy not 
exponential, and the "Lifetime" somewhat ambiguoos. 

28 -Kennedy. Khnischev and Uie Test Ban." Glenn T. Seaborg. U. Cal. Press. 1981, p. 156. See also H. 
Hoeriin. Ref. 21. 


debris.29 "STARFISH" effectively disabled or depressed operations of several sateUites. 
indicating the importance of accurate information on the intensity and distribution of 
trapped radiation for durable satellite electronics design. Information of this type on natural 
and artificial radiation has been compUed in the DARPA 'Trapped Radiation Handbook," 
which first appeared in 1971.30 The dual mission global positioning system (GPS) and 
nuclear detonation detection system (NDS) sateUites, now used for detection of nuclear 
tests in die atmosphere or m space, include a dosimeter to measure radiation in order to be 
able to estimate degradation of on-board and otiier systems as well as to detect possible 
trapped radiation from high-altitude nuclear tests.^^ 


The ARGUS concept was brought to ARPA via PSAC, as a presidential-level 
assignment, and by RYork as its first Chief Scientist York states that ARPA was die 
only place to handle the ARGUS project, and that die ARGUS idea was one of only two 
truly unique concepts in early ARPA projects. 

ARGUS was the first man-made large scale geophysical experiment in the earth's 
magnetosphere. Because of the nuclear test treaty, it is not likely that another geophysical 
expeximenr like ARGUS will be conducted again. 

A unique feature also was the role of York himself, due to his own background and 
connections with AEC, PSAC, and the DoD groups involved in nuclear testing. PSAC 
provided assistance dirough its leverage and many scientific subgroups. Yoric played die 
key role in ARPA's coordination of the entire effort; decisions were made quickly with a 
smootiily operating working group of two consisting of ARPA Uaison. CoL Dent Uy. 
who had come to ARPA from AFSWP, and die executive agent, AFSWP Chief Scientist, 
Dr. F. Shelton.32 

AFSWP. as die DoD unit concerned widi nuclear effects, had previously conducted 
several large-scale, successful nuclear test operations, but none had been of die remote, 
"task foree underway" type of ARGUS. In fact. AFSWP had just completed the 

29 -TTHJ Motion of Bomb Debris FoUowing the Siarfish Test," J. Zinn, H. Hocrlin. and AG. Petschek. 

B.M. McConnac, ibid., p. 671-692, 
3 0 -The Trapped Radiation Handbook." DNA Report 2524 H. 1971. Rev. 1973. 

3 1 ••SatelUte Verificadon of Anns Control Agreements," Chapter by Harold V. Aigo in Arms Control 
Verification^ Ed. Tsipsis, Pergamon 1985. 

32 F. Shelton, ibid., footnote 6. 


HARDTACK Pacific Johnston Island tests in mid-August, and sent a new task force 
directly from the East Coast to conduct ARGUS in the South Atlantic at the end of 
August.33 Also, many of the physical measurements involved in ARGUS, from satellites 
and remote sites, were new. AFSWP deserves much credit for ARGUS success.^ 

The IGY, predominantly an acadenaic and laboratory activity, provided an important 
assist: many preparations had already been made, including the Explorer sateUites scries, 
which provided essential and timely information on the Van Allen Belt and its 
characteristics, and many other large-scale, ground-based and rocket measurements that 
probably made the operation more acceptable and feasible than at another time. However, 
it was delicate to manage the relations between the open IGY and the classified effort 

The recorded ARPA outiay for ARGUS is about $9 miUion. There appears to be 
two reasons for this low figure: the AFSWP major costs were handled as part of those for 
operation HARD TACK, and the Explorer sateUites buUt at the University of Iowa by Van 
Allen and his graduate students were very cheap. The industry involvement was mainly in 
modification of die existing X-17 rockets, and supply of some others of a type already 
available. NASA also provided considerable assistance to AFW AL for its rocket project. 

During the course of the project, and before the acnial explosions, it also became 
recognized theoretically that some of the initial concern about the synchrotron radiation 
from the artificial belts may have been exaggerated, since the geometric distribution of tiiat 
radiation was limited to the high angles of die planes perpendicular to the trapped belt and 
could only affea sidelobes of missUe defense and most other radars. However, the major 
concern was the potential damage to reentry vehicles,^^ and to determine the injection and 
trapping efficiencies fiom an nuclear explosion required an experiment 

The technical and operational risks, both intrinsic and due to the extraordinarily 
tight schedule, were very high indeed, and as indicated in Hawkins' account, even 
increased substantially by ARPA during the operation. The success can be credited partly 

33 See Annex by W. Hawkins for a key paiticipam's view. 

34 A unclassified AFSWP movie "Project ARGUS" can be obtained from DNA Made shortly after the 
explosions, the results given there represent an uncertain early stage of the analysis of results. 

35 Discussion with H. York, 5/88. 


to many factors such as the high quality of technical effort, and the incentive of clear top 
level interest, but largely must be described as just very good luck.36 

ARGUS' impact was mainly answering, in a timely fashion, a top policy-level set 
of questions then considered highly important Also, even though ARGUS was conducted 
with a limited scientific instrumentation, it has left technological data of enduring value 
regarding trapped electrons in the earth's magnetosphere injected by nuclear explosions. 
These data have been used in design and assessment of manned and unmanned space 
vehicle vulncrabiUty. in the design of the GPS/NDS system, and in recent SDI studies. 
However, the U.S. high-altitude explosion STARHSH appears to have been conducted 
without enough preparation, due partiy to the lack of a strong foUow-on program after 

36 Dr. Shelton states that ARGUS' success was due to the "right people being in the right place 
right time," Cf. footnote 6. 






















AO 4 












... AFCRL * 



• ••»•••• 





I- 3 





-ysTARFisH [• IL 



SPACE RAD.**.. , _ 

W I0N05 









Willis M. Hawkins 
Lockheed Company, Burbank, CA 


Willis M. Hawkins 

With some detailed exploration I can pinpoint the dates of the Argus Adventure. It 
was late 1958, Lockheed's fledging "MissUe Systems Division" had emerged four years 
earlier to continue the development of pilotless aircraft exemplified by the X-7 Ramjet test 
vehicle and the Q-5 Mach 3 target drone, both of which were originated in the advanced 
design department of the California Aircraft Division. It should be remembered that the 
start of the Missile Systems Division paraUeled the beginnings of the Air Force and Navy 
ballistic missile programs. 

This fresh, new MSD organization had made unsuccessful proposals to the Air 
Force for both ICBMs and IRBMs but in the process had suggested a means for doing 
research on reentry phenomena which was successfiiL This program produced a reentry 
test vehicle-the X-17. This was an ingenious device, a 3-stagc rocket with a large size (for 
its day), the first stage with fixed fins, a second stage, with conical skirt, made up of a 
cluster of three 9-in. dia. specially-developed rockets, and a third stage using one of these 
new rockets, also with a conical aft skirt, and on the nose a 9-in. simulated reentry body 
heavily instrumented (sec Fig. A-1). The tests consisted of launching the vehicle leaning 
out to sea (Pt Mugu) a bare few degrees. After first-stage burnout die complete vehicle 
reached apogee at approximately 600,000 ft and staned falling back to earth. stabiKzed by 
the fixed fins as the atmosphere became dense enough. From this stabiUzed position the 
2nd and Bid stages were fired reaching Mach numbers near 15 at altitudes not much over 
10.000 to 20,000 ft, simulating heat input of a reentry body. Data were transmitted to 
shore before and after the "blackout" The Air Force fired about fifteen of these test 
vehicles (called the FTV-S Series) and the Navy fired approximately 20. These programs 
were pertinent to the Argus because there were five of these test vehicles left over from the 
Air Force and Navy programs when ARPA, under Herb York at the time, decided to 
confirm the trapped radiation theories of Dr. Oiristofolis. 


The total program was conceived to fire a nuclear device at high altitude in the 
South AUantic and to measure the characteristics and propagating paths of the resulting 
radiation with approximately 20-30 sounding rockets and a satelUte. Although the program 
was held in the tightest security, various instrumentation sutions were alerted around die 
work to record perturbations from nuclear events. 

Lockheed first started on the program approximately May 8th of 1958 and the team 
set out to prepare for die nuclear devices, modify the X-17's to be stable when all three 
stages were fired in an upward direction, create instrumentation for the sounding probes, 
and prepare the launching ship to be deployed in die South Atiantic off Tierra Del Fuego. 
ARPA contracted for die probes Uirough the Air Force Weapons Ub (a young officer, to 
become General Lew Allen, was the Scientific Director) for die AFWL project and ARPA 
contracted direcUy widi Lockheed a diink) for die nuclear launch vehicle.i Lockheed 
responded widi a combination of Dr's. Martin Walt and George Taylor for the science 
aspects and Tom Anderson supported by Tom Dudley widi Irv Culver (die designer of die 
X- 17) for the engineering and hardware. 

ARPA and die science community were told tiiat it would take three launches to 
guarantee one success and we were off and running widi die five spare X-17s as a 
resource. The test vehicle was long and slender and would have to be spun to be suble 
after its first stage fins were lost at separation, so one of die vehicles was prepared for a test 
of strap-on spin rockets and die structural becf-up calculated to strengdien die attachments 
between stages to make die bird widistand spinning. The fins were also canted to produce 


Thanks to die schedule, die launch stand for die diree vehicles on die ship fantail 
(The U.S.S. Norton Sound) had to be tackled first so die ship could leave to reach its 
launch station. Dudley tackled diis while Anderson and Culver tackled die spin and 
stnicmral integrity. The Air Force was charged widi die transport of test vehicles and 
nuclear devices to rendezvous widi die ship. Simultaneously, probe rockets were being 
assembled from where ever diey were available and instrumented by Walt while Taylor 
woiTied about die nuclear device furnished by Sandia. The momenmm built instantiy and 
our first fUght from near Port Hueneme was hoped to be just a confirmation. Not so! The 
bird Gong and slender) spun up just right so diat its rpm matched first bending frequency 
and we scattered hardware all over die Pacific. We had one spare left so we attempted to 

Actually, the contract was through OMR's field project branch. 


avoid disaster by additional bccf-up and reducing the fin cant With our last spare we fired 
from the fan tail of the ship to test one of the launchers and the structure at the same time. 
Disaster again, and no more birds and no time-the ship had to leave. In the time for the 
Norton Sound to reach the South Atlantic we scoured the country-tried to find smaller spin 
rockets-designed and built new str^n fittings, etc. We had to get a special courier fiighi 
from the Air Force to take the new hardware to match up with test vehicles and ship. We 
also prayed a bit 

At this point the scientific community, forgetting that the reason for three test 
vehicles was to get one success, then asked for three different altimdes. Some hard words 
were said by Tom Dudley who was in charge of launch details on the ship, supponed by 
Dr. Taylor, who then thought about it and decided to use some Kentucky windage (his 
hobby was building and firing ancient Kenmcky rifles) so he launched, or attempted to, on 
the roU of the ship (which was substantial in the weather encountered) in order to vary the 
altitude. The vehicle without spin rockets (final configuration) performed adequately from 
a mechanical and structural standpoint but its stability left something to be desired. 

Miscellaneous otiier victories and problems ensued, but the first launch on Aug. 27, 
1958 reached a still arguable altitude witii a successful nuclear event On Aug. 30 the 
second launch reached a different altitude and also fired. Finally, die maximum desired 
altitude was reached on Sept 6 witii the tiuid nuclear event The multiple teams at 
WaUops, Puerto Rico and Cape Canaveral launched probes, die Air Force read out 
experimental packages on coordinated Adas launches and Dr. Van Allen, who had 
monitored everytiiing, obtained further information from Explorer 4. All told, it was a 
triumph for science, a remarkably successful engineering accomplishment and a 
monumental logistics miracle. Science, industry and government all did it right under 
ARPA-this is the way we need to do it today. 

There are two amusing postscripts. Communications were necessary to alert 
everyone when the launch took place (under high security) so coded messages were relayed 
via misceUaneous foreign and U.S. commercial ships to die United States.^ It appears to 
be a fact that the launch trigger for the probes was via a Greek ships captain. 

The second postscript involved security. The day after die last shot. Bob Bailey, 
the P2V Program Manager from die Lockheed Aircraft Division, called me from Tahiti 

2 The Navy Task Force had been shadowed by Russian trawlers, but these were "lost" during a storm in 
the Caribbean. 


where he was on vacation. His words were "WiUy - what the hell are you blowing up in 
the South Atlantic?" I was stunned and asked him what he meant. The circumstances he 
described involved a group of instnimentation specialists alerted by the Air Force 
Geophysical Organization to listen for potential signals from the shots. They were 
discussing the whole affair in a bar where Bob and his wife were having a cocktail and 
someone mentioned Lockheed, which alerted Bob. He couldn't resist caUing me. I was at 
the time Assistant General Manager for the Missile Systems Division so he surmised that I 
was involved. So much for security. 

The whole operation started in May and was over early in September-- 
appioximately 90 days. I hope DARPA can guide us through many more miracles like 




The TIROS (Television and Inftaicd Observation SatelHtes) project involved active 
orchestration by ARPA of concepts and capabiUties into the design of a meteorological 
sateUite experimental system, the funding of the first such system together with its launch, 
and provision for foUow-on analysis, before transfer to NASA. TIROS, the first dedicated 
meteorological satellite, opened up a new meteorological era. There has been a lasting 
impact since TIROS and its successors: TOS (TIROS Operational System) ITOS 
(Improved TIROS Operational System) and, more rccentiy, TIROS N, 30 satellites in all, 
have been the principal global operational meteorological systems for the U.S. While used 
primarUy for weather forecasting and climate research projects by mAA and NASA, 
TIROS data and technology have been useful for the design of the Defense Meteorological 
SateUite System (DMSP). TIROS also provides data dircctiy to miKtary meteorological 


By die spring of 1958 there was considerable evidence Uiat technology had 
advanced to the point where it appeared possible to develop and construct a meteorological 
sateUite, and a lot of enthusiasm to actuaUy do it. The concept of using sateUites for 
meteorology had been discussed in die U.S. since the late 1940*s, and developed in some 
detaU in a 1951 RAND report by Greenfield and KeUogg.^ The International Geophysical 
Year (LGY) included plans for a meteorological sateUite. Several payloads, brought to high 
altitudes by rockets, had taken large-scale picmres of cloud patterns. "Introduction to Outer 
Space." a publication issued by die President's Science Advisory Committee (PSAC) in 
March 1958, summarized current views:^ 

1 S.M. Greenfield and W.W. KeUogg. "Inquiry Into the Feasibility Readier R^ni^s^^ From a 
SateUite Vehicle." Rand report 1951, reissued (unclassified) as Rand Report R.365. Aug. 1960. 

2 Quoted in J. R. KiUian, Sputnik Scientists and Eisenhower, MTT Press 1977. p. 94. 



Figure A-L Nuclear Warheads Were Launched Into Space by X-i7s Under the 
Auspices of "Project Argus/* These Missions Were Carried Out Aboard the 

U.S.S. Norton Sound. 


The satelUte that wiU turn its attention downward hol(^ great pron^ for 
meteorology and the eventual improvement of weather forecastmg. Present 
3cr sStions on land and sea can keep only about 10 percent of the 
atmosphere under surveillance. Two or three weather satellites could 
include a cloud inventory of the whole globe every few hours. From this 
inventory meteorologists bcUeve they could spot large storms (including 
hurricanes) in their early stages and chart theu- directions of movenwnt^^ 
much more accuracy than at present Other instrumente m the sateUites will 
measure for the first time how much solar energy is f^g on the earth s 
atmosphere and how much is refracted and reflecte|d back mio space by 
clouds, oceans, the continents, and by the great polar icc fields. 

TTicse predictions were largely fulfilled with the first few TIROS satellites. In May 
1958 Roger Warner, of the ARPAADA staff, set up a comminee on meteorological 
sateUites. chaired by W.W. Kellogg of RAND and including representatives of the three 
miUtary services, the Weather Bureau. NACA^ and RCA. TTiis committee went to work to 
define a sateUitc meteorological system and develop solutions to the many associated 
problems. The program objective was:* 

To test experimental television techniques leading to a worldwide 
meteorological information system; to test sun angle and horizon sensor 
systems for spacecraft orientation; to obtain meteorological data for rescaicn 
and development analysis. 

The committee recommended cloud cover observations using cameras of high, 
medium, and low resolution, and measurements of the earth's radiation in tiic infrared. 
RCA had participated in the early RAND study and Air Force surveillance sateUite studies, 
and since 1956 had been working for the Army to develop a system (JANUS) to be 
launched by an Army rocket to provide a reconnaissance capability.^ A prototype satellite, r 
JANUS n. was constructed, but was long and thin, without directional stabUity. About 
this time, however, the Air Force was given responsibiUty by H. York, then DDR&E, for 
all DoD satelUte surveillance systems. ARPA also had requested die Army to develop a 
booster. JUNO H based on die JUPITER, to put larger sateUites in orbit. This allowed 
RCA to modify its design to a spin-stabUized "hatbox" shape. The TIROS project and the 
name originated in die ARPA meteorological committee. Invoking an urgent requirement 
for a meteorological sateUite to assist operations of optical surveUlance sateUites, ARPA fell 

3 NASA was established later, in July 1958. 

4 In "Meteorological Satellites," Library of Congress Staff Report for the Commiuee on Aeronauucal 
and Space Sciences, U.S. Senate, March 29, 1962. 

5 -A Preliminary History of the Evolution of the TIROS Weather SateUite Systems." by John H. 
Ashby. NASA, 1964. p. 10. 


that TIROS offered a timely opportunity to reorient the RCA effort toward meteorology, 
which would not have as stringent optical resolution requirements as demanded by 
targeting/suiveiUance systems, and so could be accompUshcd with systems that were 
considerably smaller and lighter. 

This also allowed the TIROS project to be unclassified, which for a number of 
reasons was considered highly desirable at the time.s By July 28 ARPA Order # 10 was 
issued for a "Meteorological Payload" TIROS, providing nearly $8 milUon to the Army 
Material Command, under which the Army Signal Corps R&D labs were responsible for 
the payloads. with RCA the contractor J Only one payload launch was called for in the 
RCA contract The Air Force Cambridge Research Laboratory was given nearly $1M for 
data analysis in ARPA Order # 26 of September 29, 1958. The Air Force Systems 
Command was provided $3.6 million for Thor vehicles for the TIROS launch, on April 10, 
1959. OriginaUy TIROS was to include an optical television system, the top priority, to be 
bmlt by RCA under Signal Corps supervision, and an infirared scanning (IR) system built 
by W.G. Stroud, of the Signal Corps laboratory, but this IR system was not included in 
the first payload. 

When the TIROS project was transferred to NASA on April 13, 1959. the project 
plans and funding for initial payload construction, launch, and data analysis were in place, 
as weU as apportionment of responsibility in each of these areas. According to a 1962 staff 
report for the Committee on Aeronautical and Space Science of the U.S. Senate on 
meteorological satellites:^ 

The TIROS program, originated by the Advanced Research Projcc^s^f 
of the Department of Defense, was transferred to NAS A on ^pnl 13, 1959. 
Basic responsibUity was apportioned as follows: U.S. Army (U^a:>kui- 
and contractors from industry-primarily RCA); development of payload 
and selected ground equipment, data acquisition, and data transmission; 
U S Air Force (HMD and contractors from industry-Space Technoiop 
Laboratories. Douglas and Lockheed); development of launch vehicle, 
mating of vehicle and payload, launch data acquisition. Air Force 
Cambridge Research Center assists witii data analysis and mtcrpretanon. 

6 There were strong pressures to define systems to be taken over by NASA. ^ • 

satfiUite, offered much pubUc appeal, and international goodwill opportunity. The transfer to NASA 
included provision to supply TIROS data to DoD. 
RCA has built TIROS systems ever since. 

8 "Meteorological Satellites," ibid., footnote 4. 


U.S. Navy (Naval Photographic Interpretation Center) assists the WeaU^er 
Bureau in locating photographed areas by identifying landinarks and other 
geographical features. NASA (Goddard Space Hight Center), overal^ 
direction and coordination, tracking and orbit prediction operation of ^^^^ 
control center, data analysis, and interpretation. U.S. Weather Biffeau 
Oargcly Meteorological Satellite Uboratory, which is supported by NAbA). 
Dato analysis and interpretation, data dissemination, and histoncal storage 
of data. 

The same staff report gives a chronology of related events which occurred rapidly 
in diis period. For example, Vanguaiti H, carrying a dual photoceU system for earth albedo 
measurement designed by die Signal Corps R&D Lab. was launched in February 1959 to 
fulfill U.S. IGY commitments. IGYsmdies leading to the Vanguard payload had explored 
many of the aspects of a meteorological sateUite system. ARPA was aware of these studies 
and the Signal Corps lab's capability through this project Explorer VI was launched in 
August 1959, carrying a payload which transmitted a rough picmre of the earth's surface 
and its clouds. Also, in August of 1959 an Atias missile carried a camera which took 
pictures of clouds over the Caribbean and the South AUantic. And in October 1989, 
Explorer VH carried IGY instruments to measure the earth's radiation balance.^ 

TIROS 1, however, was the first dedicated meteorological satellite. A description 
is given by the same sta£f report:i° 

TIROS I (Television Infrared Observation Satellite) 

Date of launching 

April 1. 1960. 

Launching vehicle 

Three-stage Thor- Able adapted. Liftoff weight, over 105.000 pounds; total 
height, 90 feet; basic diameter, 8 feet 

General shape, weight, and dimensions of spacecraft 

A "pillbox," 42 inches in diameter and 19 inches high, covered by solar 
cells with three pairs of solid-propellant spin rockets mounted on baseplate. 
Shell composition: aluminum alloy and stainless steel. Total spacecraft 
weight, 270 pounds. 

9 "Meteorological Satellites." ibid. 

10 ibid. 


Spacecraft Payload: 

Instrumentation: Two television cameras that are identical except for lens 
equipment - a low resolution and a high-resolution camera - both with 50U 
lines per frame and a video bandwidth of 62.5 kilocycles; a magnetic tape 
lecoider for each camera with maximum capadty of 32 photographs taken at 
30-second intervals (while out of ground-station range); two timer systems 
for programming future camera operations as set by a program command 
from either Fon Monmouth or Kaena Point sutions; sensing devices for 
measuring spacecraft attitude, environment, and equipment operation. 
Antennas: four rods from baseplate for transmitters and one vcmcal rod 
from top center for receiver. Transmitters: TIROS broadcasted its picture 
on two FM radios at 235 megacycles with 2 watts each and trackmg 
information on 108 and 108.03 megacycles, with 30 miUiwatts. Power 
supply: nickel/cadmium batteries continuously charged by 9,200 solar 
cells. Power output average about 19 watts. 

The TIROS orbit was nearly circular, at about 500 km and with an inclination of 48 
dcg. Ground command of the cameras aUowed control power savings; readout was also 
commanded from the ground. 

TIROS I was an instant success. Designed for 90 days operation, in 78 days it 
provided approximately 19.389 pictures of die cloud cover which were considered useable, 
and also some pictures of the sea ice useful to ice reconnaissance." The TIROS low 
resolution, wide-angle camera TV system provided most of the data. The infrared scanning 
system was not included in TIROS I.^^ but infrared horizon sensors were employed. 
Some of these pictures showed features which were immediately identified as hurricanes 
and tornados. While routine daUy worldwide data widiout interruptions was achieved only 
in 1966. TIROS has been considered semi-operational from the first launch." Teams of 
meteorologists were involved in analysis of TIROS data, which were used to correct 
weather maps (see Fig. 1) for control of missUc firings at test ranges, and for hurricane 
tracking. The comparison of vortical cloud images and of predicted vortical structures on 
weather maps was particularly striking.^* That the payloads of the subsequent TIROS n, 
m and IV, between November 1960 and February 1962, included only minor changes of 
die television system indicated soundness of basic design. These later TIROS systems also 
included infrared scanners, radiometric and earth radiation balance measurement systems. 
Figure 2 shows the evolution of the TIROS system to 1978. The 30 satellites of the 

1 1 "Meteorological SaieUites," WiUiam K. Widger, Jr. Holt. N.Y. 1966, p. 136. 

12 Ashby.Ref.4.p.37. Infrared Sensors were used to deterniinc the horizon. 

13 Footnote S, p. 126. 

1* -TIROS Meteorology," by Arnold H. Glascr, AFCRL Report 613. 31 Mar. 1961. 


mOS. TOS and ITOS and NOAA series launched in 1985 aU included vidicon TV 
systems' similar to that of the first TIROS. The TIROS series has been the principal global 
operational meteorological system for the U.S. Weather Service.15 Beginning with TIROS 
IX, the subsequent operational meteorological satellites, other than the GOES 
gcoschronous weather satellite, have all had polar orbits. 

1 5 A c^Kn^nf "Hinhnl Weather SaieUitcs-Two Decades of Accomplishmeni." presented at the Aviaiion 
Sp^Vrf;er? ?ol^^^ and "25 Years of Weather Satellites." RCA Engineer. 

VoL 30, August 1985, p. 23. 


As the TIROS data began to be assimilated, the limitations of TIROS coverage due to its 
fixed spin axis, dependence on solar illumination, and the location of ground command 
stations began to be appreciated. In fact. TIROS was able to produce images of less than 
25 percent of the earth's cloud cover. However, this was far more than available before. 

It was soon clear that military requirements for detailed cloud conditions at specific 
times and locations would not, generally, be met by TIROS or any civilian system. 
Designs began for a military meteorological satellite system.^^ The statistics of cloud cover 
provided by TIROS and its follow-ons, as well as the TIROS system technology, have 
been important inputs to the design of the military system.^^ 





via Gt<tc*Anofi 





« •< 

I tl 

Tines i 
(A#tttc 'ita 



1 3 

« ( 
« • 



(jAMu**v ti;ai 



: .'«aAA I 


locroik* tirn 



.HOaa i ^CA%f 
m •< ! * ' 






A I 

AOVANCf 0 rtiio$.N 


«iscui Misiom 

Figure 2. From A. J. Schnapf, Ibid. 

IDA TE-214, by R.S. Warner, Jr.. Dec. 15. 1959. 
n Discussion with C. Cook, 12/79. 

The resulting Defense Meteorological SateUite Program (DMSP), in operation since 
the mid.l960s (and also built by RCA), employs two spaced satclUtes in polar orbits at 
about the same altitude as the early TIROS, with sensors covering the visible and infrared 
spectral regions, and radiometric infrared systems at different wavelengths to measure 
atmospheric structure.i8 The primary emphasis of DMSP has been on cloud cover. 
Because of technology similarities and rising costs in the TIROS sateUites and DMSP. 
Congress has questioned die need for both TIROS and DMSP. 0MB and the National 
Security Council have smdied the possibiHties of commonaUty, some of which has proved 
feasible. Also, TIROS' orbits were lowered, in the early 1970's, to more nearly that of 
DMSP.l^ However, die mUitary and civilian requirements are different, and die two 
separate systems have continued to be launched and to operate. 

Data fawn the TIROS-type sateUite are integrated with DMSP and other information 
in die Air Force Global Weadier Central at Offuti AFB. Since 1972 DMSP data have been 
available to civilian weather services. 


The TIROS idea was formed in an ARPA commiticc convened to define a satellite 
system to meet an urgent meteorological requirement related to die efficient use of 
surveillance satcUitcs. TIROS drew on previous Air Force swdies and Army technology 
developed originaUy for surveillance purposes! The top-level decision tiiat surveiUance 
would be an Air Force responsibUity made die Army^eveloped technology available for 

Roger Warner, a gifted member of ARPA's staff, puUed togedier, in die agenc/s 
TIROS steering committee, a group of experts from RAND, government labs and agencies, 
academia and industry who in fact were boUi uniquely qualified to define die system and in 
a position bodi to share and carry out die responsibility for constructing it and making it 

No new component technology needed to be developed, and the experts on the 
committee had been anxious to get going for some time. The ICY had also recommended 
such a project It would have been inefficient and unwise not to take quick advantage of 

18 "What s THe Weather Down There." by M.D. Spangler. Wesunghjuse Engi^^^^^^^ 

1974. and "Evolution of the Operational Satellite Service 1958-84- by A. Schnapf. RCA. 1979. p. i J. 

19 -We^iher Satellite Costs Have Increased.-." GAG Report RCED 86/28 Oct 31. 1985. p. 97 


this capabUity and enthusiasm. RCA. the industrial payload contractor involved, has 
continued to construct the TIROS payloads to date as weU as the military DMSP satelUtes. 
The objective to quickly obtain and use an experimental system was achieved very 
efficicnUy and quickly. After TIROS was transferred to NASA, arrangements continued to 
ensure availabiUty of data to the mUitary. ARPA can be credited with getting U.S. weather 
sateUite technology under way, which transformed meteorology, as weU as producing, 
even while in an initial experimental phase, useful information for military operations. 

Probably TIROS, or something similar, would have gotten under way eventually as 
a NASA program had not ARPA undertaken it However. ARPA's actions were on a scale 
and quality to get TIROS off to a very good start Timeliness for miUtary users, and the 
existence and nature of the accompUshment as a international interest item, evidenced by 
Presidential level announcements, were very important factors for U.S. posture in the early 
space days, and very helpful to NASA's early image. 

TIROS, however, could not be depended on to provide specific data for military 
requirements. This, plus TIROS* success, led to the development of the Air Force DMSP 
satelUtes with primary emphasis on cloud cover, as was that of the first TIROS, Negative 
lessons, such as TIROS limitations in coverage due to fixed orientation, scan angle and the 
location of ground stations, and die positive contributions of statistical information 
produced by TIROS on cloud distributions, were also essential to design the DMSP 
system. Again, tins information would have been available, presumably, if NASA and not 
ARPA had undertaken TIROS, but again timeliness would have been an miportant factor. 
Uter versions of TIROS added IR sensors. The DMSP design also incorporated similar 
technology, and DMSP data became available for civilian use in 1972. 

As a result of a 1973 study mandated by Congress, NOAA and the Air Force were 
directed to coordinate future efforts for new polar satelUte designs. However, the different 
requirements for the miUtary and civiUan users have so far justified separate systems.20 

The recorded ARPA outiay for the first TIROS was about $14 million-$9 milUon 
for payload, $4M for a booster, and $1M for analysis. Much of the development of the 
satellite package had already been accompUshed in the previous Army-funded woric, and 
die Air Force also paid for some of die expense of the ground stations involved. Costs of 

20 GACibid. 


the civilian TIROS and foUow-ons ait estimated as approaching one-half billion. The 
DMSP system cost to date is estimated also as about one-half billion 21 

21 C. Cook, ibid. 





















ARPA was responsible for getting the world's first global satellite navigation 
system (later called TRANSIT) started, with a timely, substantial push of the original woik 
at the Applied Physics Laboratory in the fall of 1958.^ The TRANSIT navigation system 
has provided reliable accurate positioning for the Navy's Polaris strategic submarines and 
other ships since the mid 1960's (fully operational in 1968), A commercial version served 
more than 8(X)0 users in 1986 including more than 20,0(X) ships and a large number of oil 
drilling rigs at sea. The system's surveying capabilities (the reason for the name 
TRANSIT), accurate to a few meters, have contributed to improvement of nearly two 
orders of magnimde in positioning accuracy on the earth's land maps including those 
generated by the Defense Mapping Agency. TRANSIT is scheduled to be replaced by the 
DoD Global Positioning System (GPS) which uses different technology, in 1996. 


In the section about ARPA in his recent autoWography, Herbert York, ARPA's first 
chief scientist, says TRANSIT was the only Navy Space proposal at the time.^ York also 
says that most of the things ARPA touched in these early space days had, in fact, been 
around a while. TRANSIT, however, had only been invented in March 1958, about the 
same time that ARPA began. When the Sputniks were launched in late 1957. researchers at 
Johns Hopkins Applied Physics Laboratory (APL) found that by accurately measuring the 
time-varying doppler shift of die radio signal from Sputnik as it went by, they could 
determine its orbit, and McClurc of the same laboratory suggested that this procedure could 
be inverted: from a knowledge of the satellite orbit and the doppler measurements, it was 

1 ARPA Order #25 of 9/25/58 to BuWeps. DcpL of the Navy, for $8.9 million for a "doppler navigation 

2 R Yoik, Making Weapons. Talking Peace. Basic Books. 1987. p. 146. York points out diat the idea 
of using a cooperative satellite for position location was old. However, obtaining the informauon 
from doppler measurcmenis and the equation of motion in a gravitational field was new. 


possible to detennine the location of the measurement.3 The sateUite orbit could be 
determined by ground stations and communicated to the satelUte. which in turn could 
transmit updated orbit parameters to the "user." along with the cw signal for doppler 
determination. VWth a computer, die "user" could quickly determine his location. 
There were some striking advantages over otfierSonns of navigation:* 

1 Since the measurement of angles or directions are not required, single 
nondiicctional receiving antennas suffice. Directional antennas aboard a 
rolling, pitching ship are complicated and create a serious maintenance 

2 Since optical measurements are not involved, the system would be immune to 
the vagaries of the weather. For months on end, the skies over the northern 
Pacific and Adantic Oceans are cloud covered. During such penods. celesual 
navigation is useless. 

3. All of the equipment sites that are required to operate the system could be 

within the U.S. This avoids the poUtical and logistical problems associated 

with operating stations in foreign countries. 
4 On land, repeated Doppler "navigation" at a fixed site becomes a new form of 

surveying. The earth could be surveyed globally in an internally consistent 

coordinate systeoL 

These features were particularly attractive for use by a submarine, which could 
briefly expose a sinaU antenna at suitable tiines, to quicidy determine its^^^^^ 

Within a month after the analysis of the fint doppler measurements:^ 

... "the essential elements of the present day Tiansit System "^^^f^^ 
in a 50-page proposal to the Navy Bureau of Ordnance complete with blocJc 
diagraii. power and weight estimates, and an accuracy analysis.. 

Although the Navy was then engaged in developing the Polaris system, and gave 
informal support to the woric at APL. apparenUy some in the Navy did not want to say an 
improved navigation capabUlty was needed at that time. Because ARPA then had DoD 

3 "TTie Gestation of Tninsii as Perceived by One ParticipanC by y-^yati, Johns Hop^s.^1^^^ 
ibid. p. 3. John Hopkins APL Technical Digest, Jan. - March 1?81 Vol 2 # 'P;^' 3^^^^^^ °; 
the Technical Digest is dedicated to TnmsiL Cf. also "The Genesis of Transit, internal APL memo by 

G.C. Weiffenbach, Mar. 1986. , ^ , ^ • i 

4 "SateUite for Earth Surveying and Ocean Navigation." H.D. Black, ibid p. 3 Cf. ^ J^!^^ 
U^"pianetary AppUcaiions of Navigation and GeodeUc SateUitcs^- by John D. Nicd^dcs. Marie 
M^omLTan^m M. Kaula. Advances in Space Science and Technology, Vol. 6. 1964. 

5 T. Wyatt, ibid., p. 32. 


rcsponsibiUties for sateiUtes. APL brought their transit proposal to ARPA in early fall of 
1958. ARPA responded positively in October with funding and authorization to build 
spacecraft and ground stations and soon afterwards for launch vehicles. The scope of 
woik program included most of the elements eventually in the operational system (see Fig. 

1 . Spacecraft (always called "satellite'* whether in the shop or in orbit) - design, 
construction, and operation; 

2 . Tracking stations design, construction, and operation; 

3 . Injection station - design, construction, and operation; 

4 . Navigation equipment - design and construction; 

5. Geodesy - expansion of the then-current knowledge of die earth's gravity 

6. Launching vehicles - design, construction, and field operations after the first 
few launchings. 

injoction Statibn 

Transmits new orbKat 
'ptnwntfm and' time oorrwtion 

Coinptttin9.C«w«« Hv'ij^^^ • 

' ■ U. pWMMtMV>and^ti0iB;!OOf>scuon . .. 

stgdtKt4-^^^M^'''''^'^^M^'^-^'^^ ^. ■■■■■ 

/rwardi.and fS^tim Ooppltf itgrah^ ^ 

t« refraction cometad Oopptar dm - ^ 

Pmonm data: . ■ ^j^^'^-;.;';- 

■ - longitude; and:tim0^-s^^^;igvj 
. corraaion ^ ■ ' :-^^'^'^^'y ^ 

Figura 1. System Architecture of the Navy Navigation Satellite System (Transit). 

From H.D. Black, Ibid., p. 4. 

6 T. Wyau. ibid., p. 32. 


The APL projea engineer states:*^ 

in May 1959, APL issued a program plan identifying an ARPA 
experimental phase and a Navy operational phase. The plan optmnstically 
envisioned six launchings in the foUowing fiscal year and eight more m the 
subsequent two years to achieve a fuU operational capability in 1962. i ne 
planincluded design and manufacture by APL of launch vehicles (possibly 
based on an adaptation of the Polaris missUe), a worldwide complex ot lo 
ground stations, and 18 shipboard navigating equipments. 

I accept full rcsponsibiUty for the design of a plan so wildly ambitious. 
Only slighUy less astonishing than the plan, however, was us ready 
acceptance (including its estimated cost) by the Depaitment of Defense. 

Soon afterwards, however, DoD assigned aU miUtary launch responsibUity to the 
Air Force. Arrangements for the launch vehicles were then made by ARPA on the basis of 
the evolving TRANSIT payload characteristics, the developing launch vehicle capabUities 
and availabUities and. the needs of other "piggy back" payloads.8 some of tiiese other 
payloads included an NRL radiation experiment (GREB), a Naval Ordnance Test Station 
(NOTS) package to measure infrared background, and the Army Map Service's SECOR 
radio location package, to permit determination of its comparative accuracy. The early 
TRANSIT satellites (one version shown in Fig. 2) were all built by APL. These evenmally 
weighed about 1 10 lb of which most of die additional weight over 50 lb for die working 
system was for redundancy and other safeguards. Arrangements for die initial launchers 
were made expeditiously: Seven vehicles, at a cost of - $28 million were provided for by 
ARPA between 4/59 and 7/59.^^ The Air Force THOR-ABLE and THOR-ABLE STAR, 
each capable of launching several hundred pounds into the required - 1000 km orbits, were 
used for the first launches. This orbit was to be nearly circular and far enough above the 
earth's atmosphere to avoid appreciable drag. The IGY Baker-Nunn satellite tracking 
cameras were helpful in determining early orbits. 

The first TRANSIT launch was in 1959. WhUe this launch faUed to achieve orbit, 
it stiU provided useful dopplcr data. The next TRANSIT, IB, achieved orbit in 1960 and 
demonstrated feasibiUty of die system. Three more TRANSITS, of evolving design (sec 

7 T. Wyatt. ibid., p. 32. Transit launches supported by ARPA were: one m 1959 which ^f^J^^^;"^ 
OTbk^but provided useful doppler from dkta: one in 1960 which achieved orbit and demonstrated 
feasibUity; ihice in 1961; and two in 1962, of which one was for Geodesy. 

8 See IDA TE 205 of 12/4/59, "Revised Development and Fundmg Plan for TRANSIT." by Roger S. 
Warner of IDA/ARPA staff, which ouUines ihe history and plans to that date. 


Fig. 3),io were launched in 1961. The first TRANSITS were not oriented and had nearly 
omnidirectional antennas. Two frequencies were broadcast in one circularly polarized 
mode to allow compensation for ionospheric effects. Later TRANSITS were smaller, used 
unfolding solar cell frames, and eventually were gravity-stabilized toward the earth's 
center. This allowed directional antennas to be used, decreasing power demands. The 
move to smaller satellites was planned in order to make use of the less expensive SCOUT 

Center support tube 
Command receivers 
Command logic 

Oewar flask and oscillator 
Command system battery 
Magnetic damping rods 

Upper radiation shield 
Memory system 

Terminal board 
Nyton lacing 
Outer iacing ring 


1 ft 


between center structure 
and outer lacing ring 

Lanyard guide tube 

Antenna coupling network 

Auxiliary nickel cadmium battery 

Telemetry system 

Battery voltage 
sensing switch 

Main and SECOR 
nickel cadmium 

Oewar fiask 
and oscillator 

54-324 MHz 
multtpiier amplifier 

162-216 MHz 
multiplier amplifier 

Permanent magnet 

Solar cells 

SECOR transistor package 

DC converters 
Cylindrical structure 

Figure 2. Cutaway View of TRANSIT 3-B Satellite Illustrating Key Components 

(U.S. Navy and APL/JHU) 

9 Thus ARPA Order 17, Task 4 of 4/59 provided nearly $5.1M to the Air Force for a ThorDelta and 
Thor 104: Task 6. of 4/59 for two TTior HusUcis, for nearly 3.4M; and A.0. 97 of 7/59 for TTior Delta. 
TTior 104, and Thor Agena; all for launches of navigation satellites. 

10 John D. Nicolaides, ibid., p, 168. 

11 Roger S. Warner, ibid. The solid propellant SCOUT was a NASA developmcnL The 1^^^^^ 
SCOUT is described in "A New Dimension." NASA Reference Publication 1028. Dec. 1978. p. 704 


It soon became dear that geodetic knowledge would have to be improved in order 
to attain the desired accuracy for POLARIS, and that this knowledge would have to be 
developed largely by experiments with TRANSIT itself. 

T mniii 

l42-7IUf 1 


IMS finuiNCT Mosi tf -r sn. ui-ita 
M uuiEB ti mam *r sn, u-ntm 

MiiKwr. rni 
: ctfion 



20it Ml 




mj Ml 

I7S.1 ^ 




1 tub* 


U] ha 




1 ia70te 

1 971 ka 

m ka 

1 iniha 






j iUM. 

• Note: Xav IB ceased radiating on July II, 1960. 
Xav 3B cntcrctl the atmosphere on*h 30, 1981. 
Ijiunchcd pickaback on Transit 4 B. 

Figure 3. TRANSIT Satellites Launched During i960 (U.S. Navy and APUJHU). 

From NIcolaldes, Ibid., p. 176. 

According to an overview by the project enginccri^ 

the number and the variety of satdUtes ultimately found necessary were 
not anticipated at the outset It was assumed in the first program plan mat 
50% of the sateUites would be launched and operated successfiiUy and that 
successful satellites would have an average life of one year. No allowance 
was made for mistakes or for the extent of the design evoluuon. 
Unfortunately, these assumptions were overly optimistic, tariy on, u 
became evident that the Transit program would require special-puipose 
sateUites for geodesy, radiation measurements, radioactive «otope power 
supply trials, and attinide-control experiments. Some of these satellites, ot 
course, had as their primary missions the support of national objectives 
ote than Transit Therefore, the number of APL-built f tcUites d^^^^^^ 
partially related to the Transit program grew to a total of 36 by *c nmc me 
system was declared fully operational in October 1968. Eight of me 
sateUites were victims of launch-vehicle faUures and two were damaged by a 
high-altitude nuclear test (Project STARFISH). 

12 T. Wyatt, ibid., p. 33. 


The STARFISH event took place in 1962, after ARPA involvement in TRANSIT. 
In fact, some of the early TRANSIT satellites have had useful lifetinaes of over 10 years.^^ 
Geodesy, in particular the accuracy of models of the earth's gravitational field, was soon 
found to be a limiting factor to TRANSIT, It was not until about 1965 that a model becanje 
available allowing die desired < 1/4 nmi positional accuracy for POLARIS. 

The fiist POLARIS submarine was declared operational by the Navy in late 1960. 
By 1963, some operational use was made of TRANSIT by POLARIS; in 1968 the 
TRANSIT system was declared fully operational by the Navy.^^ The system was not 
adopted by NASA, however, possibly because of its inability to track geostationary 
satellites, 15 Commercial use of TRANSIT also dates from 1968. The commercial 
Magnavox receivers use only one frequency, and also use a simplified cycle counting 
technique possible with reception of signals from an entire pass of the satellite. Receivers 
on Navy ships use two frequencies to allow ionospheric compensation and more 
sophisticated algoritiims which use only a segment of a single satellite pass. DMA, for 
mapping purposes, has developed its own receivers. 

The current TRANSIT system consists of a constellation of about seven satellites 
and a ground tracking network. The Navy plans a phascout of TRANSIT in about 1996, 
when the GPS, which does not use the doppler principle, is scheduled to be available. 
GPS is to provide global, real time navigational fixes of higher accuracy than TRANSIT. 


The TRANSIT proposal was brought to ARPA by APL, a major contractor- 
operated R&D laboratory of Uie Navy. While the original nootif was scientific curiosity, the 
implications of die TRANSIT concept were quickly ^predated at APL, which also had 
responsibilities for the POLARIS project 

Apparendy die Navy would not suppon the proposal at the time. To demonstrate 
feasibility could be expensive and risky. Panly, die risks were Uiose of a new space 

1 3 An account of TRANSITS successes and problems are given by Thomas A. Siansell, Jr. of Magnavox. 
in "The Many Faces of TRANSIT." paper presented at the 38ih meeting of the Institute of Naviganon. 

1* Joint paper on the Navy Navigation Satellite System (TRANSIT) Stauis and Plans." by O.L. 

Scntman, Robert J. Danchick. and Lawrence J. Ranger. APL 1987. 
15 "Technical Innovations in The APL Space Department," by R.B. Kcrshner, APL Technical Digest. Vol 

#4, OcL 1980, p. 264. 
1^ Kershner, ibid., p. 26S. 


system with a very high premium on reliable, accurate performance at a time when launch 
reUabiUty was not high and there was Uttle experience with reUability of space systems. 
While the key principle involved seemed straightforward and had already been checked, 
roughly, using the Spumiks, and no major new technology development appeared to be 
necessary, it was not clear at the outset that the accuracy of better than 1/4 nautical mile 
needed for POLARIS, could be attained, A number of experiments with the system were 
needed to develop a much improved model of the earth's gravity field before this accuracy 
was demonstrated. 

ARPA responded very quickly with funding in sufficient quantity to cover 
construction of the satellites and related ground stations, plus several launches and support 
systems, for an ouUay of about $28M at this stage. This was enough to give TRANSIT a 
very good chance of getting through a feasibiUty demonstration. ARPA bought the APL 
development plan and gave tiiem a ficce hand, except for arrangements for the launch 
vehicles and added payloads-which ARPA did itself. TTiis enabled APL to concentrate on 
the sateUite and ground system. Regarding the ARPA management die APL project 
engineer states: 

The work at APL was also facilitated by die rapidity witii which d<«isions 
could be obtained from a streamlined DoD organization. Durmg the ttrst 
year. Roger S. Warner. Jr. (of ARPA) was both the pomt of contact and the 
decision maker. In the foUowing year or two. the entire DoD management 
team comprised only two or duee individuals. The government s program 
managers were both highly competent and highly motivated. 

WhUe there was some POLARIS support from 1959. Uiere was some difficulty in 
obtaining adequate Navy funding through 1961. ARPA funding in 1960 and 1961 for 
TRANSIT appears to have been about $24M. for a total outlay of about $42M. The 
strength of ARPA support, rapidity of progress, demonstration of feasibUity. and 
diminishing expected costs ensured Navy support from 1962 onwards. It took until about 
1965. and an expenditure by die Navy in the hundred miUion range, to achieve the accuracy 
desired for POLARIS. By diis time die POLARIS budget was high, so diat tins was a 
small fraction. 

ARPA also made TRANSIT known to odier potential military users, such as DMA, 
and also in the civilian maritime area. The impact of TRANSIT on mapping, geodesy, and 
land surveying were somewhat anticipated and have been very great. An unanticipated, 

17 T. Wyau, ibid., p. 32. 


major impact occurred in oil rig placement in ocean shelf regions.i^ The impact on 
oceanography has been very great. 

About 36 operational TRANSIT sateUites have been launched, at a systems cost to 
the Navy approaching $1/2 biUion. The commercial invesmaent for TRANSIT navigation 
equipment has been estimated as about $1/2 billion.20 whUe the GPS system, now 
scheduled to replace TRANSIT (and other DoD navigation systems) by 1996, uses 
different technology, the success and reliabiUty of TRANSIT may be credited with 
establishing the basis for wide acceptance of a satellite navigation systein.21 

18 Satellite Doppler Tracking and its Geodetic Applications." Phil. Trans. Sociegof Unto 
A.294. 1980, pp. 209-406. An account of a discussion on this topic held at the Royal Society lO-l l 
October 1978. 

19 Thomas A. Stanscll, ibid., p. 93. quotes Dr. Ewing, head of Columbia Universit/s Lamont 
Laboratory, to this effect, regarding the dcvelopmcm of oceanography. 

20 Discussion with T.A. Stanscll, 1^0. 

21 Discussion with Dr. C. Cook, 12/89. 




















* ^ ^ 4 LAUNCHERS 

A097 J 



























CENTAUR, the first Uquid hydrogen-Uquid oxygen burning upper stage for 
efficiendy placing sizeable payloads into geosynchronous orbit, or into lunar and deep 
space missions, was first funded by ARPA in 1958. Transferred to NASA in late 1959, 
CENTAUR, after a number of problems and faUurcs. had its first successful orbital flight 
in 1963, and its first successful mission in 1966. Since then it has been a very reUable 
"workhorse" for placing payloads, including DoD's FLTS ATCOM, into geosynchronous 
orbit. A version of CENTAUR is planned to go on the Air Force's TITAN IV. 
CENTAUR engine technology has also been used in the upper stages of the large 
SATURN rockets used in the APOLLO manned flight series to the moon (see Chapter V), 
and in the liquid hydrogen-oxygen engines also used by the SHUTTLE. 


The advantages of a hydrogen-oxygen fuel combination to achieve high exhaust 
velocities were recognized by early rocket pioneers. U.S. efforts on liquid hydrogen 
propulsion systems date back to before WWH. at NACA^s Lewis Flight Propulsion 
Laboratory. The engineering difficulties of the necessary cryogenic systems were 
recognized during WWH in the U.S. and Germany. After WWH the Air Force funded 
work on liquid hydrogen-liquid oxygen (LH2A-OX) fueled rockets at Ohio State 
University, and some fundamental woric in the same direction was conducted at the NACA 
Lewis Laboratory. Those early experiments showed that exhaust velocities in the range of 
3500 m/sec could be attained with LH2A-OX. Early smdies of sateUites, including some 
directed to achieving orbit with a single stage, recognized the potential advantages of an 
LH2/LOX combination, particularly if housed in light, internally pressurized structures.^ 
In this 1945-1950 period some significant earlier studies of figures of merit of different 

1 Notably the Martin HATV vehicle design, studied for the Navy's B"«f of Aeronauucs. John L. 
Sloop. "Uquid Hydrogen as a Propulsion Fuel." 1945-59. NASA SP 4404, 1978. p. 44. 


vehicle weight and propcUani combinations in the U.S. and Germany, were further 
extended.2 However, these early initiatives were not followed up immediately. 

A number of major advances in engineering large-scale liquid hydrogen generators 
and storage systems were made by the Atomic Energy Commisson (AEC) in the early . 
1950s for their early work on thermonuclear devices. In the mid-1950s also, foUowing 
recommendations of their Science Advisory Board and of NACA's Lewis Laboratory, tlie ' 
Air Force commenced efforts to use Uquid hydrogen for aircraft propulsion at -high 
altimdes. This work led, in 1955, to fUght tests of a Lewis-designed jet engine in a 
modified B-57 aircraft. Soon thereafter the AF commenced the (then) classified project 
SUNTAN, in which Pratt and Whimey (P&W) was funded in the 1956-1958 time period 
to develop an LH2-buming engine for a high-altimde surveiUance aircraft envisaged as a 
successor to the U2.3 SUNTAN took advantage of much of the AEC-developed LH2 
technology and made a number of further advances, notably in pumping LH2. Evehmally , 
(in 1958) P«fcW successfully ran an LH2 turbojet engine with ratings approaching the 
desired surveillance aircraft's characteristics. SUNTAN was dropped in 1957. however, 
partly because of controversies over the surveillance range capability the LH2 technology 
would allow, but mostly because, after Sputnik, attentions turned to satelUtes for the 
surveillance mission. 

About the same time, K. Ehrickc of Convair made proposals to the Air Force for an 
LH2-fucled upper-stage system based partly on Convair's tiiin-skinned, pressurized 
structure technology used successfuUy in the Atlas inissUe. Pratt and Whimey was also 
proposing, together wiUi Lockheed, die appUcation of die LH2 technology lessons learned 
in SUNTAN to upper stages to boost large surveUlance satellites into geosynchronous 
(GEO)orbiL In July 1958. die Air Fbrcc SUNTAN management team suggested to ARPA 
(which had overall responsibiUty for DoD Space Systems) a joint Convair-P&W effort 
which would build on die strong points of bodi organizations. At die time, die IDA staff 
supporting ARPA (ARPA/IDA) included several individuals who had strong backgrounds 
in related propulsion technology.^ R. Canright. one of these experts, was involved in 
developing an early ARPA plan for launch vehicles matched to payloads including 
provisionforuscofLH2A-0X upper stages.5 NASA, which was just estabUshed. as one 

2 Notably by W. von Braun in Gennany and R. Canright of JPL. 

3 "Liquid Hydrogen as a Propulsion Fuel," ibid., p. 141. 

4 Ibid., p. 180. 

5 -Proposed Vehicle Program." IDA TE 1 10. 16 Feb. 1959. GP. Sutton and R.B. Cannght 


of its first actions, formed the Silvcrstein Committee to coordinate national plans for large 
space vehicles. Early considerations of the Silverstein Committee brought out advantages 
of LH2/LOX upper stages, and ARPA acted quickly, before the end of August 1958, to 
fund, through the Air Force, a new Convair-P&W proposal for CENTAUR with 
LH2/LOX engines to be used as an ATLAS upper stage.^ 

Soon thereafter, in October 1958. NASA requested transfer of CENTAUR, which 
was worked out the following year with Air Force continuing as manager and NASA 
promising to develop a number of CENTAUR upper stages, for which die "user" agencies 
would supply payloads, and an overall NASA-DoD Steering Committee which included a 
DoD representative widi icsponsibiUty for future DoD coimnunication satellites.'' Large 
communication satellites, in geosynchronous orWt, were envisaged as high priority miUtary 
payloads. A Uttie later, still another DoD-NASA committee made an intensive study of the 
characteristics of die large launch vehicle SATURN, reconomending adoption of the 
proposal Uiat SATURN upper stages use LH2/LOX. The Army BaUistic Missile Agency's 
(ABMA) von Braun group, which was building Uie SATURN, initially opposed LH2/LOX 
because of its dangers and the light structure involved, but eventually agreed to it* 

Reflecting early optimism as weU as the strongly felt need for its capability, die first 
CENTAUR flight test was scheduled for January 1961. 

CENTAUR was "die" rocket by which NASA would conduct extensive 
eanh orbit missions, lunar investigations, and planetary studies. Aside 
from military missions assigned to CENTAUR, which were to be 
considerable, NASA planned to launch one operational CENTAUR every 
montii for a period extending weU into die 1970s and beyond.^ 

NASA had initially assigned CENTAUR management to its Marshall Space FUght 
Center, apparently because of diat Center's responsibUity for SATURN, a much larger 
project, including die planned use of CENTAUR-related engine technology for S ATURN's 
upper stages. 

^ AO 19 of 8/58, CENTAUR, for $21 J million. 
^ Ibid., p. 201. 

8 Ibid, p. 238. 

9 "History of CENTAUR," NASA Uwis Research Center, undated, p. 2, For comparison, 11 
ATLAS-CENTAUR launch capabilities were 4-^ear. 


Figure 1. CENTAUR. This Version. Made for an ATLAS Second Stage, Is About 

9 m In Lenflth and 3 m In 

1 0 D. Baker. The Rocket. The History and Development of Rocket and Missile Technology." Crown, 
NY, 1978. p. 147. 


The CENTAUR configuration then envisaged, shown in Fig. 1, involved two 
P&W RL-10 engines with about 15,000 lb of thrust each.^i jhe nozzles, subject to the 
high temperature hydrogen flame, were also cooled by the Uquid H2. The practicability of 
doing this had been proved in previous work at several laboratories. CENTAUR was 
eventually to place more than four tons into low orbits, nearly two tons into 
geosynchronous orbit (GEO) and nearly one ton into an earth escape trajectory in 
combination with ATLAS and TITAN first stages. Figure 2 shows a typical trajectory to 
There were considerable technical issues involved: besides those of the cryogenic 
systems for the LH2/LOX fuel, there were the pumping and control of these Uquids in a 
zero-gravity environment, the embrittlement of the thin-skinned structural sections 
subjected to low temperature, the complex nozzle cooUng system, precision control of 
starting and restarting two engines, and the navigation and propulsion control systems for 
achieving precise orbits. 

These issues proved to be too much for such an optimistic schedule, and there 
ensued a stream of test stand explosions and failures. In March 1962 the first CENTAUR 
flight test exploded shortly after liftoff. These events dampened DoD plans for use of 
CENTAUR, in particular for project ADVENT, which had the objective to place a (then) 
large communications satellite in geosynchronous orbit. NASA then reassigned 
CENTAUR rcsponsibUity to their Lewis Laboratory, and in November 1963 the first 
successful (single stage) flight took place. Shortly thereafter the SATURN upper stage 
Centaur-type LH2/LOX engines were also successfully operated. 

In 1966 a successful series of CENTAUR-lifted missions began. During this 
1961-66 period diere were also improvements in the size and accuracy of computer- 

1 1 Baker, op. ciL, p. 147. Table 1. p. 167. 

12 From H.M. Bonesiecl, "ATLAS and CENTAUR Evaluation and Evolution," Convair-Gcneral 
Dynamics Co., 1982. 

13 A.D. Wheclon. "The Rocky Road to Communicaaons SaicUite." AIAA 24th Aerospace Sciences 
Meeting. January 6-9, 1986, AIAA Document 86^93, p. 5. There were plans m 1958-59, for 
several DoD communication saielllies. to be placed in GEO orbits by Centaur in 1962. IDA ^.29. 
Mar. 27. 1959. "Insianiancous Global Saielliie Communications Systems." by S.B. Batdort. xnese 
were eventually passed by ARPA and DoD to the Army's project ADVENT. Sec S AMSO ctuonolog>^ 
1954-59. Air Force Systems Command. Space Division History Office, p. 117. ™ ADVENT 
experience had many repercussions in DoD, one of which was the fonnation of the Defense 
Communications Agency, I. Getting. "All in a Lifetime," Vantage 1989. p. 534. 


controUed incrnal navigation and guidance systems. According to a Lewis Laboratory 

Coupled with already proven Adas first suges. Centaur vehicles sent seven 
Surveyor spacecraft tb probe the surface of the Moon between May 30, 
1966 knd January 7, 1968, furnishing valuable dau for the first manned 
landing on the Moon in July. 1969. 

Other important AUas/Centaur missions followed, mcluding b<»sang the 
Orbiting Astronomical Observatory to scan the stare firom above the Eaith s 
atmosphere ... sending two Mariner spacecraft to chan the planet Mars ... 
















vuwtf n eNGiNf anom 


Figure 2. Atlas/Contaur Parking Orbit Mission Delivering a Spacecraft to 

Synchronous*apogee Transfer. 

14 "Hisioiy of CENTAUR," ibid., foomote 7, p. 3. 


launching two Pioneers to Jupiter on a solar system escape trajectory and a 
Manner to Venus and Mercury. 

The Centaur stage combined with the Air Force Titan IH booster provided a 
capabiUty to launch larger spacecraft like Helios A and B around the Sun, 
two Vikings to Mars, and two Voyagers to Jupiter, Saturn and beyond. 

Centaur has flown not only exploratory scientific missions but also those 
with terrestrial benefits such as Applications Technology Satellites and the 
Intelsat. Comstar and Fltsatcom communication satelhtes. Centaur has 
delivered these domestic and military communication satelhtes into 
geosynchronous orbit 

Centaur today is a mature, high-energy, still-viable upper stage witii an 
overall operational reliability record of 96% ... 100% since 1971. 

As Centaur begins its third decade, it is being modified to fit into the Space 
Shuttle as a high-energy upper stage and wiU launch the Galileo spacecraft 
for further study of Jupiter and its moons as well as send the Ulysses 
spacecraft over the poles of the Sun. 

However, after the ChaUenger disaster, NASA cancelled its plans for use of 
CENTAUR with the Shuttle, after four years and $0.7B of effort, citing safety issues. 

The major DoD use of CENTAUR to date has been to launch FLTSATCOMS. 
Since the mid 1970s a more recent (1988) assessment credits ATLAS/CENTAUR witii 
6.75 tons to low earth orbit (LEO), and cites a new LH2A-OX engine at the top of the 
priority Ust of the focussed-technology projects now funded by the Air Force under the 
DoD/NASA Advanced Launch System projects.^^ CENTAUR is also paired with TTTAN 
IV in future Air Force plans.^^ 

Table 1 shows CENTAUR missions until 1982. Figure 3 illustrates the 
construction of the SLVD-3D, the most recent ATLAS-CENTAUR combination. 


Much of the CENTAUR technology was available in 1958 when the Air Force 
brought the Convair proposal to ARPA. The ARPA staff for CENTAUR was headed by 
R. Canright, who was thoroughly familiar with LH2/LOX technology. The key cryogenics 

15 Launch Opuons for ihe Future," Congressional OfTice of Technology Assessment, 1988. p. 5 
1^ Discussion with Dr. C. Cook, 12/89. 


Table 1. CENTAUR Missions 

(59 Missions) 

(7 Missions) 

Mission Type 


Mission Type 


Test Flight 

Applicattons Technology 

Satellites (ATS) 
Orbiting Astronomical 

Observatory (OAO) 
Manner Mars 
Intelsat IV 
Intelsat IV A 
Pioneer F 
Pioneer Q 

High Energy Astronomical 

Obsen^atory A 
High Energy Astronomical 

Observatory B 
High Energy Astronomical 

Obsen/atory C 
Pioneer Venus 
Intelsat V 


Test Flight 
Helios A 
Helios B 
Viking A 
Viking B 
Voyager 1 
Voyager 2 


Atlas/Centaur (last 36 Flights) 
Titan/Centaur (six Flights) 
Centaur Stage (last 40 Flights) 

and engine technologies had been investigated extensively by NASA and the Air Force, and 
the Ught stnicniral technology was an adaption of that used in the ATLAS nussUe. Several 
leaders in early space technology felt that LH2/LOX was needed for a variety of missions, 
especially for powering second stages to geosynchronous orbit Apparendy the only 
technical group that did not favor CENTAUR at the time was von Braun's team, which 
whUe forward in concept was conservative in its engineering. ARPA's timely action gave 

17 From H,M. BoncstceK "ATLAS and CENTAUR Evaluation and Evolution." Convair-General 
Dynamics Company, 1982. 


CENTAUR an early, substantial boost, and probably moved its schedule ahead some 
months. The effort thus staned may have helped to get the CENTAUR LH2/LOX 
technology past the von Braun group's objections, since they eventually agreed to it for 
SATURN upper stages, for which they were responsible. NASA leadership was 
convinced of the merit of LH2/LOX and undoubtedly would have pushed it anyway. There 
were ambitious early plans for CENTAUR's use, and assignment of CENTAUR 
responsibUity was made to HuntsviUe, evidendy in the belief that engineering difficulties 
had been overcome. After several failures, however, CENTAUR responsibility was 
reassigned to the group more famiUar with cryogenic engineering, the Lewis Laboratory. 
These early failures forced cancellation of ADVENT, a major joint-Service program, and 
somewhat negatively influenced the subsequent miUtary usage of CENTAUR, its main 
utility overall having been for NASA flights. However. CENTAUR has put the 
FLTSATCOM satelHtes in orbit from the mid 1970s. The degree of acceptance of 
LH2/LOX technology as efficient, economical, and practical, evidenced by die CENTAUR 
launch record indicates the coirecmess of the ARPA and NASA judgements. CENTAUR 
technology was essential for the APOLLO missions, and is used today in one of the 
TITAN rv configurations, and, with new hardware, in the LH2/LOX SPACE SHUTTLE 
engines. CENTAUR, in a variety of versions is still a "workhorse" today, and of value to 
U.S. space capability that is hard to overestimate. 

The total, one-time recorded ARPA outlay for CENTAUR was $22M. The total 
cost of CENTAURS launched to date appears to exceed $2 bilUon.^' 

18 C. Cook. ibid. 

19 C. Cook. ibid. 


Figure 3. SLV/3D Fabrication Sequence 








PRE WW tl 






1 1st CENTAUR 















¥ -I 






























"The authorization of a large rocket vehicle by the Advanced Research Projects 
Agency in August 1958 and assignment of its development to the Army Ballistic Missile 
Agency (ABMA) marked the beginning of a series of successful large launch vehicles."^ 
Besides support of the original proposal of the Von Braun AMBA group, the ARPA 
suggestion of using a cluster of available rocket engines to achieve large first stage thrust at 
an early date and at low cost proved highly successful. Together with use of the Uquid 
hydrogen technology developed earlier for the CENTAUR vehicle for the upper stages, the 
ARPA-initiated SATURN I series was used in tests for the NASA's APOLLO program and 
later for the SPACELAB program, for a total of 19 successful flights. 


There were a number of initiatives in the mid-1950's for large boosters in the 
milUons of pounds thrust range. In 1956. for example, the Air Force Science Advisory 
Board (SAB) made a recommendation for such a development This led, a Uttle later, to an 
Air Force effort to constmct a single-barrel Uquid propellant rocket engine approaching 5 
miUion pounds thrust, eventually called N0VA.2 In early 1957 the Army's BalUstic Missile 
Agency (ABMA) rocket group under Wemher Von Braun began smdies of an approach to 
a large booster involving a cluster of rocket motors.^ In late 1957. after Spumik. a more 
specific design for such a vehicle, using a cluster of four Rocketdyne E-1 engines to 
achieve about 1.2 miUion pounds of thrust, was included by this ABMA group, under the 
name JUNO V. as a major feature of a proposal for a "National Integrated Missile and 
Space Development Program." This was only one of several proposals for large rocket 

"Liquid Hydrogen as a Propulsion Fuel." by John L Sloop. N>^A SP 44^^ NASA ^^^ f^^^ 
1978. p. 223. and "Stages to Saturn" by Roger E. Bilstein. NASA SP4206. p. 23. 1980, Biisiein 
givcs a detailed technological histoiy of the Apoilo/Satum launch vehicles. 

A brief history of early U.S. rocket developments is giv«i by a »«y P^^lf^*- ^^^J^ SSS^ w^^^ 
presidential Science Advisors. George Kistiakowsky. in A Scierutsj at the White ^ouseV^zrd 1976. 
pp. 95-99. Tlie name NOVA, confusingly, was used for several different booster approaches. 
-A History of the Saturn 1/lB Launchers," by David Baker. Spaceflight 1978. p. 146. 


programs at the time, made in tlie strong national desire to "catch up and move ahead" in 
space. The ABMA proposal aimed to make available quickly and cheaply, for whatever 
national programs might be undertaken, a large booster capable of putting payioads of 
many tons into orbit It was fairly clear that a manned space program would have such a 
requirement and at the time it was believed also that large military communications and 
surveiUance satelUtes might be needed. One of ARPA's main tasks after its formation, 
largely in response to this national push, was to make rational choices among these options 
and to move things ahead rapidly."^ 

Soon after its inception, ARPA was invited to present its plans for launch vehicles 
to the National Security Council. ARPA's representatives recommended the use of clusters 
of available rockets and the use of Uquid hydrogen and Uquid oxygen (LH2A-OX) to make 
efficient upper stages.5 Remaricably prescient regarding subsequent events, these 
recommendations reflected the backgrounds and expertise of the then ARP A/IDA staff.^ 
While the idea of using clusters of engines offered the advantages of redundancy, to some 
it appeared coiiq)lex, with the possibility of difficult control problems^ 

After consideration of the Von Braun group's proposal, Canright and Young of 
ARPAADA suggested the use of a cluster of 8 MB-3 (again Rocketdyne) engines, which 
had been proven in the JUPITER and THOR programs, rather than the four still 
developmental E-1 engines proposed by ABMA. TWs change was agreed to by Von Braun 
and the JUNO V clustered booster project got under way in August 1958.« The engines, 
however, required considerable modification to be used in a cluster configuration.^ 

The first goal of the program was to demonstrate the feasibility of the engine cluster 
concept by a full-scale, captive firing. In September the project's scope was extended to 
include at least four flight tests. ARPA Order 47 provided for tests for the captive and for design smdies of funire launch faciUties. Figure 1 shows one of the early 

4 "Making Weapons. Talking Peace.- by H. Yoik. Basic Books. 1987. p. 142. ff. 

5 "Liquid Hydrogen as a Propulsion Fuel," ibid., p. 224. 

6 The National Security CouncU presentations were made by R. Canright <>f AM'A/TO A who ha^^^ 
active in hydiogen-oxygen rocket research at JPL and assistant director for missiles at McDonncU- 

^ Kistiakowsky, ibid., footnote 2. 

8 AO 14, 8/15/88, for $92.5 milUon. ^ 

9 "Stages to Saturn," ibid., p. 79. details this history and emphasizes the low cost aspect of this early 

10 AO 47 of 12/58 $8.4 million. 


vehicles returning ftom a static test In November a new, more ambitious objective was 
approved: "to develop a reliable, high performance booster to serve as the first step of a 
multistage carrier vehicle capable of pcrfonning advanced space missions."^^ 

Figure 1. Removal of the Booster From the Static Test Stand 

In February 1959, at ABMA request, ARPA approved a change of the clustered 
booster project's name from JUNO V to SATURN. The first SATURN fiight was planned 
for October 1960. The upper stages had to be chosen well before then, and an ARPA 
study of this issue in May 1959 recommended using a two-engine TITAN configuration as 
second stage, with several CENTAUR engines in the third stage. Again the motif for this 
choice was to move ahead with available and near-future technology as far as possible.i2 
However, soon thereafter, H. York, the first DDR&E, proposed to cancel SATURN, on 
several grounds:!^ (1) the only justifiable national mission for a very large booster was 
manned space flights; (2) there were no military missions that required manned space fiight 
and all justifiable military missions then envisaged could be Ufted by the TTTAN and its 

1 1 Second Semi-annual Technical Summary Report on ARPA Orders 14-59 and 4-7-59, by ABMA, U.S, 
Army Ordnance Missile Command, 15 Feb. 1960. 

12 Discussion with J.C. Goodwyn, 10/88. 

1 3 Quoted in "Liquid Hydrogen as a Propulsion Fuel," ibid., pp. 227-228. 


planned fumre extensions (in particular, several smaU communications satcUitcs. which 
could be handled by TITAN, were better than a few larger ones); (3) SATURN as then 
being constructed was not large enough for extended manned space fUghts, which should 
all be undertaken by NASA. Similar viewpoints were apparenUy held by Kistiakowsky, 
the President's Science Advisor, and his PSAC advisors.i^ However, R. Johnson, head 
of ARPA at the time, strongly maintained that there were military needs for large payloads, 
especiaUy for manned vehicles capable of maneuvering and renmiing to earth.i^ as a result 
of York's proposal a joint DoD-NAS A committee was convened to consider the by now 
multifaceted pToblem,!^ which included: (1) Defense payloads and boosters to lift them; 
(2) NASA's future need for large boosters; (3) ABMA's future, largely tied to SATURN, 
(4) transfer of ABMA to NASA.i^ This committee considered SATURN, TITAN and 
NOVA, concluding that SATURN (in retrospect SATURN 1) was die best bet for the near 
future, citing also its greater payload capability and operational flexibility. The committee 
also recommended further study of upper stages. York reversed his views, apparently 
partly as a result of the recommendation of this committee, and partly because to keep 
ABMA aUve, SATURN, its major occupation, would have to be funded inidally by DoD. 
Shordy aferwaid, ABMA was transferred to NASA. 

As the joint DoD-NAS A committee had recommended, the issue of second stages 
for SATURN was studied by NASA and ABMA. Eventually the viewpoint of NASA's 
Lewis laboratory prevailed and LH2A-OX was recommended for the second and third 
stages." nic third stage was to use a cluster of CENTAUR RL-10 engines, and for the 
second stage a larger, 200,000-lb thrust LH2/LOX engine was to be developed. Shordy 
afterward the "SATURN vehicle team" was formed with NASA and DoD participation. 

U Kistiakowsky.ibid.,p.80: "it was our conclusion that aside from poliutal conside^^ 

Se S » to S to abandon the Saturn and to concentrate on the NOVA, starting with a high 
^e nS^Ja vehfcle and giaduaUy piogicssmg to '^^^'^^i^ y^^^'^^,,^^^^^^^ 
Soviets superior to us until Se laie I96(fs,bat ensures a reasonable overall level of effort and ensures 
the space program as a truly civilian effort" 

1 5 Johnson especially had in mind "MRS V". a maneuverable returnable space vcjucle a concept in many 
ways simXmhe current project NASP. The AF was soidyuig. ^^l}^^:^^^^^,^Z^c 
hypersonic space vehicle. Not long after SATURN'S .t^^^^JJ^ ^ASA Jo^^^ 

exiuu of his considerable acuviiy in this connecuon is described m Richard J. Barber, History of 
ARPA, 1958-75, Sec. HI to m^l. 

16 Kistiakowsky. ibid., p. 75. describes SATURN as an inseparable mix of technical and admimstranve- 
political problems. 

1*' Bilstein, ibid., p. 38. 

18 Report to the Administrator. NASA on SATURN dcvelopmem plan, by SATURN vehicle team, 
IS December 1959. 


under the chainnanship of A. SUverstein of NASA, to review more closely and recommend 
definite SATURN configurations to meet anticipated NASA and DoD needs, including 

The Silverstein group recommended sequential development of a SATURN "C" 
family of vehicles, beginning with the SATURN Q, later called simply SATURN, with 
the ARPA/ABMA developed first stage, and upper stages at first based on the CENTAUR 
RL-10 engine, and later, for the second stage, the new 200,000-lb thrust L-2 LH2/LOX 
engines. Still later SATURN, acconiing to this plan, was to use a cluster of milUon- 
pound-thrust NOVA-type engines as a new first stage, together with the L-2's for the 
second stage and RL-lO's for the third 20 This "map" of tiie Silverstein committee was 
largely followed in subsequent events, through SATURN V. the vehicle for the manned 
lunar expeditions. 

On the basis of the Silverstein recommendation NASA now planned a 10-vehicle 
SATURN C-1 flight series, using the ARPA/ABMA first stage, to be followed in 1967 by 
the larger SATURN C3 (or SATURN V) type. With highest national priority assigned in 
1960, two SATURN Cl's were planned for launch in 1962. A thrust of 1.3 million 
pounds was achieved in April 1961, in a captive, flight-rated test of eight clustered H-1 
engines at Marshall Space Hight Center.21 pians for successive configurations of 
SATURN had by then progressed rapidly, including provision for recoverability of the first 
stage. The manned lunar expedition in 1967 was announced in May 1961. 

The C-1 ARPA/ABMA first stage was successfully launched in October 1961 and 
in November 1961 die first industrial contract for 20 C-1 first stages was let to Chrysler for 

1 9 The Silverstein Committee had one month to come up with its recommendation. 

20 Interestingly, the ARPA representative on the Silverstein Committee. GP. Sutton. appaenUy was 
sdU recommending further studies of ATLAS type engines. This was due apparently to the <^«ire to 
use existing systems and reduce costs; LH2/LOX in this conscrvauve ARPA approach, would come 
later LH2/LOX had been previously recommended by Canright, and was pushed successfully t>y 
SUverstein. An additional reason for ABMA's deciding to choose the wider and lighter cryogenic 
engine configuration was the bending moments for then prospective heavy payloads, such as 
DYNASOAR. Discussion with J.C. Goodwyn October 1988. 

21 A chronology of the SATURN tests is given in D. Baker. The Rocket, The History and Development 
of Rocket and Missiles Technology. Crown, NY 1979, p. 243ff. 


The 10 NASA SATURN C-1 flights included several which used clusters of the 
CENTAUR type engine for second stages and a smaller cluster for third stages and tested 
APOLLO procedures and components. Except for the failure of one H-1 engine in one of 
the flights, which was nearly completely compensated for by the control system and the 
remaining engines, all the C-1 flights were completely successful.22 jhe follow-on 
SATURN IB. with the clusters of 200.000-pound thrust L-2's LH2A-OX. for the second 
stage, and CENTAUR engines for the third stage, was used to test the APOLLO system 
and its engines, including docking maneuvers in earth orbit, through 1966. In late 1966 
the test flights of die SATURN V configuration began. 

The remaining SATURN IB vehicles were brought out of storage in 1973 to 
support the SKYLAB Space Station program and the APOLLO-Soyuz project In all, 
between 1961 and 1975. 19 launch vehicles of the SATURN I family had served to 
rehearse moon landing flights and to support manned space flight programs.23 in addition. 
22 unused H-1 engines evennially were employed as first stages of NASA's DELTA 

Since the Challenger disaster there has been renewed interest in the capabiHties and 
cost of large-payload options for the future. A recent smdy indicates that large military 
payloads into GEO are likely whetiier or not the SDI continues^^. One option being 
foUowed up in a joint AF/NAS A program is the ALS (Advanced Launch System), with 
capability somewhat greater than SATURN L 


While there were similar ideas in the ARPA/BDA staff the JUNO V (predecessor to 
SATURN) proposal was made to ARPA by the Von Braun ABMA team. Initiation of the 
JUNO V-SATURN program occurred in a time of major, national concerns regarding U.S. 
posture and capabiKtics in space, and about responsibiUties for space-related activities. It 
involved an inextricable mixnire of technical, administrative and political factors. ARPA's 

22 The incrtial guidance system used in the C-l's were planned by ABMA to mvolve components used 
previously by ABMA in JUPITER and REDSTONE, which in turn stems from the system used m the 
German V2 in WW U. ARPA insisted that ABMA also use new systems those develop^ for 
ATLAS and TTTAN. The eventual ineriial package used a stable platform evolval ftoin j^ 
woikwiAinertial components stemming from ite Bilstcin. ibid., p. 243. Discussion with 
J.C. Goodwyn, October 1988. 

23 Baker, ibid., p. 245. 

24 -Launch Options for the Future." Office of Technology Assessment. U.S. Congress, 1988. 


objectives were to be able to get large payloads into orbit consistently, for whatever use, as 
quickly as possible without excessive cost. Later, when national concerns lessened, 
opposition to this route was led by York (DDR&E) and Kistiakowsky (President's Science 
Advisor). This opposition preferred a more leisurely, but direct route to a SATURN V- 
type system, ail under NASA. Thirty years later, there arc again studies of how to get large 
military payloads into orbit at low cost, re-examining the old approaches, among others. 

While it was an adhoc system involving much available technology, SATURN I 
still required engineering the engines and tanks, and the solution of a new complex multiple 
rocket system control problem. The ABMA group was probably the most experienced and 
capable in the U.S. at the time, and best able to build and test SATURN at low cost and in 
a shon time. At the same time, die ARPA support enabled this group to keep going over 
the period of transfer of space responsibiHries to NASA. The decision to use this capabiUty 
for SATURN, and keep ABMA going as a national asset seem to have been made by H. 
York, then DDR&E, in spite of his earHer views. ARPA had backed the ABMA group and 
had York's earlier opposition to SATURN prcvaUed there might have been a significant 
delay in the NASA program. 

Besides the timely ARPA initial funding action, the ARPA technical interventions 
regarding using available engines and more modem incrnal control technology had a 
significant impact on the successful C-1 series. The ARPA early action in funding 
CENTAUR'S ongoing LH2/LOX technology probably helped considerably to overcome 
Von Braun's initial opposition to this and the associated light structures for second stages. 
The ARPA plan was to use tiiis technology gradually, using initially more conservative and 
less costly second stages, but NASA's (Silverstein's) interest in LH2/LOX pushed this 
higher risk technology further for use in all upper stages of SATURN. No doubt 
CENTAUR or something similar would have been soon funded by NASA in any case. 
However, in these early days time was very important It appears also that without the 
LH2/LOX technology die NASA moon prcrjcct could not have occurred when it did.25 

While the SATURN 1 launch series was remaricably successful, doubts remain 
about the necessity for the number of flights that acmally took place. The risk of failures, 
undoubtedly very important, was lessened by the approach of the conservative Von Braun 

25 Bilstein, ibid., p. 189. 

26 ibid., p. 336. 


ARPA's rcconied outlay for SATURN was about $93M for the rocket and $8,5M 
for a test stand, totaUing nearly $102M. NASA's outlays for Saturn were about $4 billion 














A0 10 



I A014 






























ARPA pioneered the construction of large ground-based phased array radars with 
ESAR (Electronically Steered Airay Radar). Constructed in less than two years, and 
completed in the fall of I960, the low-powered L-band ESAR immediately demonstrated 
computer control of beam steering in two dimensions, with a capability of detecting and 
tracking space objects on a par with other space surveillance systems. ESAR led dircctiy to 
the Air Force Space Tracking Radar. FPS-85, which is still operational today. ESAR's 
successful performance accelerated an ARPA program of phased array components which 
has impacted all subsequent U.S. large phased array systems. ESAR's performance, better 
than predicted, at a high but not unreasonable cost, also encouraged BeU Telephone 
Laboratories to move rapidly toward construction of phased array radars for the Army's 
ballistic missile defense projects. 


In 1957 a President's Science Advisory Committee panel and many other experts 
had pointed out the need in baUistic missile defense (BMD) and space surveillance to 
detect, track and identify a large number of objects incoming or moving at very high 
speeds. Electronic steering of radar beams in two angular dimensions, more agile than 
mechanically steered antennas, offered significant advantages for this purpose. While 
several electronically steered arrays had been built before 1958, such as the Navy's TPS 48 
and TPS 33, these did not have the large aperture and high power required for BMD and 
space appUcations and used a combination of phase and frequency scanning.! A number of 
expens were skeptical of the practicality of constructing a reliable large phased array 
system, witii the technology available, at reasonable cost. An anempt to do so by Bendix 

"Survey of Phased Anay Accomplishments and Requirements for Navy Ships," Merrill I. Skolnik, in 
Phased Array Antennas, Eds. OUver and Kniiuel, Artcch House, 1972, pp. 17-18. 


began in 1958 under Air Force sponsorship and was nimed over to ARPA in accordance 
with DoD assignment to ARPA of responsibility for advanced technologies for BMD 2 

ARPA decided to open a competition for design and construction of a large 
experimental two-dimensional phased array, with beam steering under computer control. 
This was to be tiie first array steered altogether by phase control. ARPA solicited 
proposals and selected Bendix largely because of the woric they had done for the Air Force 
and the prospects they offered of using reliable low-cost, production-type technology for 
the many components involved in a phased aiiay.3 AO 29 of 9/58 provided $15 million for 
a wideband phased array radar (EPS 46-XW 1). Woric began in Spring of 1959 and the 
array was completed in November of 1960, A 90-clcmcnt Unear phased array was 
constructed first to check out the Huggins wavc-mixing approach to steering phase control, 
and otiier techniques, such as ceramic tetrodes for transmitter power amplifiers, one for 
each broadband antenna clement^ After successful demonstration of a one-dimensional 
array ESAR was extended to fill out a two-dimensional array. Figure 1 shows the 
completed ESAR array. There were spaces for 8000 elements, but only 760 were acmaUy 
connected to transmii-receive modules for the experiments involving ESAR. This, together 
with the power limitation of the available tetrodes, made ESAR a low power system, which 
was considered acceptable for an experimental program. Computer control and processing, 
key features of ESAR, were designcd-in and built with IBM participation, witii soUd state 
components used wherever possible. An account by one of the Bendix engineers states 
that ESAR was also used to develop the techniques of "Space Tapering," using fewer 
active elements witii spacing arranged to give nearly the same sidelobes, which has since 
been used in most phased arrays.^ 

A radar textbook gives a description of the system:^ an example of an electronically steerablc array using a frequency 
conversion phasing scheme. The antenna is 50 feet m diameter. Tjiebeam 
can be scanned in less than 20 microseconds. A cluster of 25 scanning 

2 IDA TE-54. Mar 20, 1959. "Technical Evaluation of Air Force Development Plan for ESAR." 

3 Discussion with A. Rubenstein, IDA, ex- ARPA DEFENDER Program Manager. 11/87. BendU s 
performance in automobile radio manufacturing was a factor in its sctecuon. 

* A description of several of these features of ESAR is given in -Electronically Scanned Air Force 
Systems I." by Moses A Dicks, et al. Radar Techniques for Detection. Tracking and Navigation. 
Gordon and Breach 1964, p. 397ff. 

5 "The AN/FPS 85 Radar Systems." J. Emory Reed, Proc. IEEE, Vol. 57, 1969. p. 334. 

6 Introduction to Radar Systems, MX Skolnik. McGiaw HiU. 1962. p. 318. 


Figure l. ESAR 

beams, 5 rows in elevation and 5 columns in azimuth, can be generated by 
the ESAR system. A separate transmitter feeds each of the L-band penodic 
antenna elements. 

Important capabiUties proven by the experimental ESAR included multiple target 
tracking, beam formation and accuracy determination, sidelobe measurements, and 
constructional maintenance procedures.^ 

Operating ESAR for tests as its construction went along was inmiediately 
successful: even with its low power it proved possible to detect and track space objects at 
least as wcU as the other existing space survdllance systems could at the time. The ARPA- 
assigned Air Force project managers for ESAR at RADC, enthused by this success, 
proposed that the Air Force construct a foUow-on, larger high power phased array radar for 
space tracking based largely on ESAR technology. Experts from Lincoln Uboratories, 
who had a large phased array study project since early 1959. were skeptical, pointing out 
that the failure rate of the numerous conventional high power electronic mbe components 

7 J. Emory Reed. ibid. 


used might be high and lead to overwhelming reUability problems. But with DoD backing 
the Air Force proceeded with the FPS-85 phased array radar, different from ESAR in 
having separate (but adjacent) transmit and receive antennas, and in a larger number of 
elements and a much higher power level, providing for the possibilities of numerous tube 
failures by arranging for a large number of people to do replacements, and pointing out the 
graceful degradation characteristics of phased arrays, demonstrated by the success of 
"space tapering" in ESAR. The contractor, again Bendix, completed FPS-85 in 1963, and 
the expected large numbers of replacement tubes were found not to be necessary in its 
operation. After a fire destroyed the first FPS-85 in 1964, it was rebuUt in 1968 with 
updated technology and components. 

Table 1 shows the evolution of large phased array technology in the U.S. beginning 
with ESAR and briefly describes the common features, and differences, of ESAR and the 
new FPS-85 together with features of other major phased arrays.* In 1968 it could be said 

The AN/FPS-85 is the most advanced operational large computer-controlled 
multifunction phased array radar. It has a range of several thousand miles 
and can detect, track, identify, and catalog eanh-orbiting objects aiid 
ballistic missiles. This system is important to the North Amencan Air 
Defense Command's space detection and tracking system because it can 
detect, identify, and track hundreds of objects concuirentiy in a constantiy 
increasing population of carth-oibiting objects.' 

The FPS-85 quickly became part of the AF SPACETRACK System, and is still 
operational today. Because of its flexibility, a scanning program to detect possible 
submarine launched baiUstic missUcs was added, making the FPS-85 also pan of the 
current ground-based SLBM warning system.^^ 

ESAR was operated as an experimental system for several years. However, FPS- 
85, which had more advanced technology, began to provide better opportunity to test 
techniques, for desirable improvements such as techniques for wider bandwidtii 

8 Radar Technology, E. Brookner, Artech House. 1984. p. 33 1. 

9 J. Emory Reed, ibid., p. 324,. 

10 "Warning and Assessment Sensors," by J. Toomay. Chapter 8 of Managing Nuclear Operations, 
Ed., A Zraket, Broomings 1984, p. 297. 

1 1 Discussion with Major General Toomay (US AF, Ret). December 1987. 

Table 1. Chronology of Large Phased Array Technology In the U.S. 
After Kahrllas, see footnote 8. 



Date 1 



One tetrode per radiating element 
746 radiating elements 
IF phase shifting 



High power-multiple transmitters 
Separate transmit and receive arrays 
Confined feed 
Thinned receive an'ay 
Diode phase shifters 



Monopulse space feed 

Diode phase shifters 



High power-multiple transmitters 
Monopulse confined feed 

Diode phase shifters 



High power 

Monopulse space feed 

Fully filled 

Diode phase shifters 



Offset monopulse space feed 
Optical magnification reflect array 
Limited scan 
Ferrite phase shifters 



Monopulse space feed 
Fully filled 

Ferrite phase shifters 



Multiple transmitters 
Monopulse confined feed 

Varying size subarrays 
Ferrite phase shifters 


Sperry Dome 
naoar (operry) 

360 deg in azimuth; zenith to 30 deg below horizon 
C-band. 1MW peak. 5 kW average. 50 ft range resolution 

o c unitimA cAsirriy frsifnA timfi 427 DDS 

^ 9 WiUIIKi oCalWI II all TO iiiii»i ^fc* KK 

Dome-cylinder items, 6 ft diameter; confined feed 


High power-multiple transmitters 
Very wide bandwMth 
Monopulse confined feed 



Solid state 

Thinned - — . 


•Since this lUt was published. PAVE PAWS is now legaided as opcraiicmal 


The important success, and the limitaiions of ES AR, lent emphasis to a broad-based 
phased-axray component and techniques program at ARPA. Substantial efforts were made 
to develop low-cost high power tubes and phase shifters, extend component frequency 
ranges, and to find ways to increase bandwidth, to apply digital techniques, and in the 
study of antenna coupUng.i^ This technology has improved all U.S. phased array 
projects. The ARPA cross-field high power amplifier developments, in particular, later 
proved important in the development of the Navy's AEGIS phased array." 

The impact of ESAR on later large phased array efforts associated with ballistic 
missUc defense efforts was less direct, but real. According to Mr. Albert Rubenstein. 
ARPA program manager at the time, BeU Telephone Laboratories (BTL). then construcung 
the Army's Nike-Zeus BaUistic Missile Defense System, were kept closely informed about 
ESAR, and a special effort was made to completely document ESAR.M TTie BTL program 
manager, however, does not recaU any specific technical impact of ESAR." The major 
influence of ESAR on BTL seems to have been by way of encouragement or provocation: 
the fact that ESAR worked well, did not have major reliaWBty problems, was constructed 
rapidly and weU documented technicaUy. and had a known cost which was not 
unreasonably high. Also, OSD confidence in phase arrays was strongly influenced by 
ESAR's success, and strengthened the basis for OSD's insistence that die Army 
incorporate phased arrays in their BMD program. 

The BeU "History of Engineering and Science in the Bell System" gives their 
history at the time:'^ 

In 1960 BeU Labs conducted fundamental investigations of phase connoUed 
scanning antenna arrays for possible application to Misale 
Defense System. Arrays vrith their inerrialess beams would provide greater 
capabilities against tiie high traffic level threat This consideranon became 

12 For example. AO 136 of 2/60 for phased amy tube development; AO 337 
shifters. AO 345. of 4/62, multiple beams Klystron for phased '^l'- 

ElectroitaticaUy focussed Klystron, of 7/63. Codiphase digital radar. AO 74, of 4/59. and also 
©ATE 196. June 1959, by T.C.Bazemoie. ,okx- 

13 "System Design Considerations of the AN/SPY- J TMnaniuer.- by OJt. LorMt. et al.. ISUi Tn- 
Service Radar Symposium. 1972. Vol. U. p. 21. 

14 Discussion with Mr. Albert Rubenstein. IDA. ARPA Defender Program Manager m 1958-59. 
December 1987. . . 

15 Discussion with C. Wanen. 12/87. BTL, very strong technically was used to going its own way. 
Discussions with Dr. C.W. Cook, and CM. Johnson, December 1988. 

1« -A History of Engineering and Science in the Bell System." M.O. Fagen. ed.. BTL. Inc.. 1978. 
p. 431. 


one of the principal technical reasons advanced in 1963 for not proceeding 
with the tactical deployment of the original Nike-Zeus System. In Nov. 
1960 at Redstone Arsenal* Bell Laboratories representatives gave a 
presentation to ARGMA on the subject of phased arrays in a terminal 
defense... to report on the study to date and to provide a basis for a 
proposal to do exploratory phased anay work ... authorization was granted 
in June 1961 ... ground breaking (was) in March 1963. 

It should be recalled that by November 1960 ESAR had been constructed and 
successfully operated. 

In 1963, at White Sands MissUe Range, BTL constructed MAR 1, the first large 
hardened phased array dedicated to HMD, under the NIKE X program. MAR used 
different phase-shifting technology than ESAR, and had considerable difficulty with 
component reliability. ^'^ However, BTL later successfully managed construction of 
several other large phased arrays in later phases of the Army BMD program, which ended 
in 1975. The last of the BMD phased arrays of this period, the high power PAR, 
constructed by GE at Grand Forks, South Dakota, is still operational as part of the Air 
Force Space Tracking System and as a threat discrimination element in the AF ballistic 
missile warning system.^s According to CM. Johnson, Army SAFEGUARD Project 
Manager in 1970, one of the approaches considered in design competition for the PAR was 
that of FPS-85, with a separate transmitter and receiver airay. A different set of 
technologies, however, was chosen for PAR, to meet the requirements for a hardened 
system. Including a common transmitter and receiver array, and the use of a "space feed" 
with fewer transmitting mbes. gave PAR a somewhat higher power and bandwidth than the 

In the mid-1960's ARPA funded construction of HAPDAR, an S-band 
demonstration low cost "hard point defense" phased array design by Spciry, which was 
located at White Sands, and has been used for a number of years in radar beam 
management cxpcriments.20 in this same period ARPA also conducted a broad technology 

17 Ibid., p. 432. 

1 8 "Warning and Assessment Sensors," by John C. Toomay . Chapter 8 of "Managing Nuclear Operations, 
Ashton Carter et al.. Eds.. Brookings 1984. p. 296-7. 

1 9 "Ballistic MissUe Defense Radars." Charles M. Johnson (U.S. Army Safeguard System Office), IEEE 
Specmim 7, 3. March 1970. pp 32-41. 

20 AO 516 "HAPDAR." 10/63. Cf. also "HAPDAR- An Operational Phased Array Radar." by Peter J. 
Kahrilas. Proc. IEEE, Vol. 56, No. 11. Nov. 1968, p. 967. 


progrm to address the problems of hardened low-cost phased am^ ADAR: an 

Advanced Design Anay Radar Study, synthesized much of this technology, and defined an 
up-to-date phased array radar system for operation in a nuclear attack environment 22 The 
crossed-field, high power ampUfication technology initiated by ARPA had an important 
later impact on the AEGIS system. 


ES AR was an extension and acceleration, by ARPA. of previous Air Force-funded 
effort, toward a "space track" radar inherited with ARPA's space responsibilities. There 
were a number of high level recommendations tiiat phased arrays would be necessary for 
the BMD mission. It was considered a risky venture at the time, pushing the state of the 
art of phased arrays, scaling up to large size, using computers to control the system and 
process its data. Dr. J. Ruina, then reponsible for missile defense R&D under DDR&E, 
was told by Bell and Lincoln Uboratories that large 2-dimensional phased airays would be 
beyond the state of the art ESAR's history seems very contemporary: in spite of the 
experts' negative views, ARPA decided to issue an RFP emphasizing cost-cutting to fend 
off strong fears about the cost of such systems, and contracted a fast paced effort to a firm 
relatively new in the game. 

ESAR was very successful, at every stage of construction and testing, causing 
considerable excitement in the RADC managers. ESAR pioneered "space tapering" and 
"airay thinning" and demonstrated the important graceful degradation characteristic of 
phased arrays. Because of the degree of high-level interest, timing of these achievements 
was critical. The same office at RADC which managed ESAR for ARPA took over 
direction of die FPS-85 with Gen. J. Toomay as program manager. Indirectiy, ESAR's 
success encouraged a major phased array effort to get going, for BMD, by BeU Labs, Bell, 
however, used different technologies in a painful learning experience. 

The ARPA phased array components and techniques program, which intensified 
after the success of ESAR, had a very broad impact on subsequent military phased airay 
efforts, and more directiy its results were used in the construction of the HAPDAR low 
cost demonstration array at White Sands, and the ADAR phased array smdy and 

21 For example, AO 136of 2/60 for phased array tube devdopmimt; 

shifters, AO 345. of 4/62, multiple beams Klystron for phased arrays; AO 436 for High Power, 
electrostatically focusscd Klystron, of 7/63. 

22 AO 498, 5 13. of 10/63, and 663 of 10/65. 


development In the opinion of several experts, this broad phased array technology effort 
was the only one of its kind, and the results have influenced all other major phased array 
efforts since that time.23 

The recorded outlay for construction of ESAR and its testing, and also including the 
early experimental work extending bandwidth using the FPS-85, was about $20M. ARPA 
outlay for the phased array technology program appears to have been about $25M. The 
original FPS-85 cost about $30M, and its replacement after the fire, about $60U^^ The 
BTL phased arrays built for the Army's BMD project cost nearly $1B. 

23 Discussions with Dr. M.I. Skolnik and Major General Toomay. 

24 Discussion with MG Toomay. 1/90. 







TPS 40 


TPS 33 

arpO army industry 


















I 85 

ft!*?^ .RESULT 










AO 32 
RFP ' 




































In response to an 18-month assignment from DoD in late 1960 to answer the critical 
question of the utility of infrared (IR) satellite early warning systems against ICBM's, 
ARPA initiated and directed project TABSTONE (target and background signal to noise 
experiments), TABSTONE was the most comprehensive and well-coordinated program of 
IR field and laboratory measurements, analysis, and technology development up to that 
time. At the end of 18 mondis TABSTONE had progressed far enough for ARPA to give a 
positive answer which raised the level of confidence in DoD and enabled development of 
the technology of the current U.S. IR satellite early warning systems (SEWS). The 
TABSTONE scientific results also had a major impact on design considerations for 
subsequent developmental progranis leading to current U.S. systems, to improvements 
(such as the Advanced Warning Systems), and to SDI programs. 


When U.S. ballistic nsissilc programs began to get under way in the mid to late 
1950's, ground-based observational systems for tracking took advantage of the intense 
light emitted in the early launch phase, and such phenomena as reflection of solar radiation 
from the plume and missile body at higher altimdcs. Soon, efforts began to measure 
quantitatively die intensity and spectral content of this radiation, some using high altitude 
aircraft. The Inter-Service Radiation Measurements Program, coordinated by the Air 
Force's Cambridge Research Laboratory, was one of the major efforts of this type. In die 
late 1950's die AF had also formed plans for infrared sensors for missile launch detection 
in its early 1 17L satellite program.^ 

In the late 1950's also a PSAC panel under William E. Bradley conducted a broad 
review of the problem of ballistic missUe defense. The panel recommended funher 
investigation of the utility of infrared and optical sensors for the detection and ffacking of 

1 Deep Black by William E. Burrows. Random House, New York. 1986, p. 84. 


ballistic missiles in the various phases of flight, including reentry. In this same time frame 
the Air Force had started studies of project BAMBI (Ballistic Missile Boost Intercept), 
which included IR sensors in space to detect and track ballistic missiles and warheads. 

At its beginning ARPA was given, by the President and DoD. broad responsibility 
for space systems. After sorting out the various miUtary satelUte proposals, ARPA 
recommended that the multifunction, complex and expensive Air Force 117L satellite 
program be divided into several simpler systems. One of these new systems was an 
infrared sateUite to detect missUc launches, named MIDAS (MissUe Detection Alarm 
System.2 jhe other systems were the SENTRY, later the SAMOS sateUitcs, dedicated to 
surveillance, and DISCOVERERS, fOT satellite technology development Responsibility 
for all these 1 17L programs, which were in advanced stages of development, was returned 
to the Air Force by KL York after he became DDR&E in 1 959. 

MIDAS was reviewed by ARPA in 1959 and 1960 and a number of 
recommendations for changes were made, mainly toward more background 
mcasurcments.3 While there were some background measurements made for MIDAS the 
program seemed predominantly target-detection oriented. These recommendadons seem to 
have had Uttle initial impact, however, and the first MIDAS tests began, in near-earth polar 
orbit, in 1960.^ 

In 1958, in response to the Bradley Committee recommendations. ARPA's project 
DEFENDER began smdies of sensors and measurement systems in the radar, IR and 
visible spectral ranges needed to improve understanding of the phenomenology of ballistic 
missiles from launch to reentry.^ Under DEFENDER studies also were conducted of 
sensors for BAMBI. some of which were infrared systems.* BAMBI's emphasis was on 
midcourse intercept, but it also required launch-phase information. The DEFENDER IR 
effort also included fundamental work such as IR emissions from flames and the properties 

ARPA HMD Technology Program Review. 3-14 August 1959. Vol. ffl (declassified) p. 1019 Air 
Force IR reconnaissance sateUite studies apparently began in 1956. 

An IDA/ARPA team review made the review. Discussion with L. Bibennan, EDA. 1 1/87 and IDA-T- 
E-157. by R.S. Warner, 19 August 1959. See also ARPA 1959 review, p. 1052. 
History afStnuegic Defense, by RX. Maust et al. SPC report SPC 742. Sept V^^^ "^^^^ 
SDd Astronautics. An American Chronology of Science and Technology in the Exploration of Space. 
1955-60. by Eugene M. Emme, NASA 1961, p. 147. 

Lincoln Laboiaioiy toolc on a major responsibility for carrying out reeiitry 'n^si^"*^'^^^"^^!? 
1960 but was not strong, at the time, in the infrared area. ARPA helped lay out the early reentry IR 
measurements program. Discussion with R. Ziildnd, 1 1/88. 
AO 6 AFSC. Task #7, 1/59. This task also included launch phase investigations. 


of molecules.7 Airborne IR measurement capabUities were considerably augmented.8 
Infrared phenomenology associated with nuclear explosions was also given attention.^ 

In the late 1950's also, significant efforts had been made by the U.S. infrared 
community which had, earlier, begun the important series of Infrared Information 
Symposia, and to make IR "state of the art" reviews. ARPA helped focus this effort by 
funding the publication of the fiist Handbook of MiUtary Infrared Technology.i^ 

In 1960 there was a review of missile launch detection programs by PSAC and 
other high level DoD committees. The main focus was the question of whether a MIDAS- 
type satellite IR system was workable. Available data seemed insufficient and unreliable. 
Recommendations were made by these groups that a new, coordinated national program be 
established to provide a better scientific basis to answer Ais important question.^ 
Additional concern regarding this question came from early reports that MIDAS satclUtes 
and some other satellites carrying related infrared sensors aU had a large number of false 
alarms. 12 An early theoretical analysis of the false alarm problem Qater shown to be 
incorrect) indicated that it might be insoluble.i^ An editor of Aviation Week described the 
status of concern: 1^ 

In the spring of 1961 the new administration's Defense Secretary, ^ot^S, 
McNamara, pubUcly expressed doubts over Uie feasibiUty of the J^AS 
concept during Congressional hearings. "There are a number of highly 
technical, highly complex problems associated with this system, 
McNamara said. "The problems have not been solved, and we are not 
piepared to state when, if ever, it will be opcrationaL" 

7 AO 6, Task 13, 4/59, At about the same time iheie vwis increased NASA research on radiation hwiing 
by rocket exhausts, cf. Handbook ofl/^ared Radiation From Combustion Gases, NASA SP 3080, 
1973. p. iiL ^ . ,K 

8 AO 6, Tasks 15. 4/59. 20. 5/59. and 31 of 4/89: the last for a "Global Sysgm to be 

1962." AO 30 of 10/58 enabled AFCRL to undertake a laigc program ($12M) of ^J!'^^'''^^ 
rocket plumes and transmission from aircraft, and "pigsyback" on ^^^Z^'^^a'^S^^ 
propeUants and aircraft measurements of backgrounds. An amendment to the A^A oi^^^J^ 
TO^SSrr satelUtc provided for a smaU NOTS sensor for background measurements espcciaUy of 
reflected sunlight for high clouds to supplement MIDAS. IDA TE 157, ibid. 

9 AO 1 1 1 of 1 1/59. MIDAS was to have some capabiUty of nuclear explosion detection, cf ., ARPA 
1959 review, p. 1024, 

10 AO161of6/60. TTie IEEE proceedings of Sept 1959 was also dedicated to a state of the art review of 

1 1 Discussion with R. Zirkind, 11/88. 

12 Discussion with R. Legault, IDA. 10/88. 

13 This analysis was made by P. Cuichis of IDA. Discussion with J. Jamieson, 12/88. 

14 Secret Sentries in Space, by PhUip J. Klass. Random House. New York. 1971, p. 175. 


The basic problem, beyond unreliabiUty troubles that then plagued aU 
satellites, was that the infrared sensors could nristake the infrared radiauon 
from sunlight reflecting off high-altitude clouds for rocket-engine pluines. 
This meant that a MIDAS sateUite passing over the USSR might mistake a 
cluster of high-altitude clouds basking in the sunhght for a mass ICBM 
attack and flash a false alarm back to the U.S. 

MeanwhUe. the Air Force was proceeding with the next phase of MIDAS, 

involving somewhat higher orbits. 

Even as McNamara was testifying, the USAF was readying two full- 
fledged MIDAS satelUtes for launch and much would be ndmg on men 
sucMSS or faUuit. The MIDAS payload weighed roughly 2,000 pounds, 
including delicate infrared sensors and complex electronics^d was 
mounted in the long nose section attached to the Agena. A powerful Atlas 
first stage was required to launch the MIDAS into the 2,000-m^e near polar 
orbit that would be needed for operational use over the USSR to give the 
spacecraft sensors a wide-spanning view. On July 12, 1961, MIDA5>-3 
was successfully launched into orbit, with an apogee/pengee alumde ot 
roughly 2,100 miles and an inclination of 91 degrees, from Vandenberg 
AFB, Calif. 

The USAF disclosed that MIDAS-3, as weU as MIDAS 4 which went into a 
similar orbit on October 21. would be tested against missUes flrcd from 
Cape Canaveral and Vandenberg, as well as against special flares designed 
to mimic the infrared characteristics ("signature") of rocket engines. It was 
shortly after the MIDAS-4 launch that the Kennedy admimstranon dropped 
the heavy security cloak over the reconnaissance satellites, and u enveloped 
the MIDAS program as well. But from informed observers it was learned 
that the MIDAS was still encountering the same problem of positive 
identification of missiles and false alarms. It was clear that much more 
experimental data, and testing, were needed to devise sensore which could 
discriminate lockct-engine plumes from sunlight bouncing oft clouds. 

DDR&E Harold Biown assigned ARPA the task of answering the question whether 
a MIDAS-type system could work in late 1960, requiring an answer in 18 months.i^ The 
TABSTONE program was set up by ARPA in response to the DoD assignment, with R. 
Zirkind as director. TABSTONE was to go back to fundamentals, and would include a 
very broad range of field measurements, many of unprecedented quality, together with 
analysis of the results, and involved a substantial fraction of the expertise of the IR 
community. As a national program, TABSTONE was able to obtain ready cooperation and 
top priority on Service assets. After a preliminary internal assessmem of the problem a 
meeting of experts was caUed in late 1960 to help define the program. 

15 Ibid., p. 176. 

1 6 Discussion with R. Zirkind, 1 1/88. 


In early 1961 TABSTONE programs got under way." The work was earned out 
by industry, academic groups, the Air Force Cambridge Research Laboratory. Navy 
Laboratories, and IDA, and also included participation by Canadian and U.K. groups, all 
under TABSTONE direction. Many of the available capabilities and ongoing programs, 
including the IRMP. were extended, and some were modified. The capabilities of a 
number of IR measuring instruments were extended and improved, and a new IR inaaging 
vidicon constructed. Chemical, physical, and aerodynamic problems connected with the 
phenomenology of IR emissions from rocket plumes at different altitudes were also 
addressed. Field measurements of missile plumes were made, some at ground level, but 
mainly from high altitude aircraft, and also from other rockets and "piggy-back" systems 
onboard the same missiles being measured, and from satellites. Measurements were made 
at wavelengths from the infrared through the ultraviolet, with as high spectral resolution as 
possible and with careful attention to caUbration. Theoretical calculations were made of die 
emissions and absorption of molecules and of rocket exhaust phenomena. Properties of a 
wide variety of propcUant compositions were measured, on a laboratory scale and in the 
field, mainly in static ground level experiments. The possibilities of countcrmeasures were 
also explored. 

Background measurements were made from aircraft and balloons. Some statistical 
information on background was also obtained from instruments on satelUtes and high 
altitude probes. Transmission measurements were made from aircraft, some using solar 
emissions, and also using long tubes containing controlled gas mixtures. 

Transmission data were analyzed in detail by a group at the National Bureau of 
Standards Boulder laboratory. These data formed part of the basis of later computer 
models of atmospheric transmission. Results on target emissions and background were 
summarized in a series of B AMIRAC (BaUistic Missile Infrared Analysis Center, set up 
under DEFENDER) reports for TABSTONE. 

Some of die TABSTONE measurements in die early launch phase contributed also 
to the BAMBI studies. TABSTONE also made some measurements in midcourse, useful 

17 AO 237 of 5/61 to ONR; AO 238 to AFSC. and AO 243 lo Navy's BuWcps, all f 5/6 L AO 236 of 
6/61 provided for the University of Michigan's BallisUc Missile Radiauon Informauon Center 
(BANflRAC) and AO 250 of 6/61 provided for NBS to collect and analyze transmission data. 


to BAMBI, but the B AMBI intercept requirements were generaUy more stringent in space- 
time resolution than those for TABSTONE.^^ 

TABSTONE results and plans were coordinated and reviewed in a series of 
meetings throughout the project, notably the yearly AMRAC meetings. TABSTONE data 
and analysis had a major impact both on understanding the early MIDAS results and on the 
subsequent developmental efforts toward infirared warning satelUtes. The TABSTONE 
results were considered sufficient, at the end of 18 months, to understand the main 
quantitative features of signal and background noise and some of the characteristics of 
filters to obtain better signal to noise. In briefings at that time by the ARPA program 
director to PSAC and to DoD. a reasonable scientific case was made for the eventual 
operable utiUty of properly designed IR warning saielUtes. Some uncertainty remained, 
however, untU the mid 1960s, and TABSTONE continued to provide important 
infonnation to its end in 1965. A symposium was held on its results in that year.i^ 

After TABSTONE had helped raise DoD confidence in IR for missile launch 
detection, the Air Force conducted related measurements programs, some using sateUites.20 
A critical review of all existing information in 1967 affirmed the continued value of 
TABSTONE data and outlined areas where further work was needed.2i In the late 1960's 
further experiments and development of a new infrared satelUte system got under way. In 
the early 1970's the Air Force's geosynchronous-orbit satellite early warning system, 
(SEWS), including IR scanning sensors, became operational.22 The present system 
includes three satellites in geosynchronous orbit, one over die Atiantic and two over the 
Pacific areas, including, besides IR warning sensors, systems for detection of nuclear 

Following TABSTONE, DARPA work in support of infrared strategic warning 
technology had a short hiams. DoD and ARPA reviews in 1968 estabHshed objectives for 
a new ARPA Plume Physics program which got under way in 1970. Theoretical models of 

1 8 B AMBI was evcntuaUy terminated for other reasons having to do with complexity and cost 

19 Communication from Dr. A. Flax. IDA, 2/90. J. Missile Defense Research, classified issue. Vol. 4 
#1, 1966, contains a preUminary review of the TABSTONE results and further references. 

20 History of Strategic Defense, ibid., p. 24. The subsequent Air Force BR satellite program was managed 
by the Aerospace Corporation. 

2 1 Discussion with Dr. H. Wolfhard, IDA, 1 1/88. 

22 "Warning and Assessment Sensors," By J. C. Toomay. in Managing Nuclear Operations, by C, Zraket 
and A. Carter. Brookings 1983, p. 306. and Aviation Week, Feb. 20, 1989. p. 34. 

23 Senate Appropriations Committee. Department of Defense Appropriations, FY 1975. parti, p. 514. 


the flow and radiation from launch and reentry plumes were formulated in this penod. 
These were further developed by NASA and the AF into standard computer models, which 
were validated to a considerable extent by experimental data under the DARPA IREW 
program in the mid 1970's. Attentions turned in the late 1970's to measurements and 
theory of high altitude plumes phenomena, appUcations of new infrared technology to 
detection and tracking of plumes and other targets, and improvement of lifetime and 
rcUability of space-based IR systems. SDI has contributed substantial support in these 
areas since its inception. 


The motif for TABSTONE was the very strong high level interest in obtaining some 
10 minutes or so extra warning time beyond horizon limited radar, by using an infrared 
satellite. There could be greater overall confidence in a warning picture developed by both 
microwave radar and infrared, which involved different physical phenomena. The Air 
Force IR MIDAS satellite was a very large program, on which ARPA's brief span of 
management had little initial impact However, MIDAS experienced severe difficulties, 
which led to its cancellation. Doubts were pubUcly expressed by Mr. McNamara, then 
Secretary of Defense, whether any such IR system could be made to work. Some 
controversy continued, however, with the Air Force's Gen. Schriever contending that 
MIDAS could have been successful.^ 

TABSTONE was set up as a national program, under ARPA management, to go 
back to fundamentals to obtain an answer to the infrared satellites question, with an 18 
months time limitation. TABSTONE was managed direcdy by an IR expcn on ARPA's 
staff, R. Zirkind, and involved orchestration of existing technological capabilities and 
making improvements where necessary to achieve a coordinated IR measurements effort of 
unprecedented quaUty. The infrared conamunity, in academia, industry and government 
laboratories apparentiy sensed the crisis caused by the MIDAS simation and cooperated 
fully. TABSTONE appears to be still regarded by this community as an IR measurements 
program of unique quality and brcadtii.25 The data obtained from TABSTONE was 
carefiiUy archived and is apparently still used by investigators in the IR area.^* 

24 Discussion with Dr. J. Ruina, 6/89. 

25 Discussion with Drs. J. Jamieson and H. Wolfliard, 1 1/88. 

26 Dr.H.Wolfhard,ibid. 


TABSTONE achieved its objective in that results at the end of 18 months were 
good enough for ARPA to give, with reasonable assurance, a positive answer to DoD on 
the question of eventual workabUity of an infrared satellite, and continued to provide 
important information for OSD decisions on IR warning satelUte systems, to the end of the 
project in 1965.^^ By this rime also there was some relaxation of concern about the 
"missile gap," due to a recent information coming from the first suiveillance satellites.^s 
This plus the construction in the early 1960's of the 440L OTH missile attack warning 
system were "stop gap" measures, while further Air Force-developed IR infrared satellite 
programs were cairied out and used to make measurements. TABSTONE can be credited 
with raising die level of confidence in DoD which led to a subtained effon toward 
developing the technology of the present DoD operational IR warning system, of its 
continuing improvements such as the Advanced Warning System, and possible future 
systems such as SDI's BSTS .29 

The recorded ARPA ouday for die TABSTONE program up to 1965 was about 
$1 8M. The SEWS system cost is estimated as about $5 billion to FY 1988.^0 

27 Discussion wiih R. Zirkind. 7/88, and A Flax, ibid. 

28 IClass. ibid., p. 176. 

29 Aviation Week, ibid. 

30 DoD Authorization Hearings before the House Armed Services Commiiiec for FY 1984. R&D. 
p. 1304. 










DARPA has developed much of the technology of high energy lasers (HEL) and 
has supported construction and test of state-of-the-an systems for military R&D, such as 
the ALPHA chemical laser. Most of this technology has been transferred to the SDI 
program. The DARPA effort also had significant impact on moderately high power lasers 
now used in industry, on the lasers used in the DoE Inertial Confinement Fusion (ICF) and 
Atomic Vapor Laser Isotope Separation (AVLIS) programs, and on the materials and 
components in lower energy lasers used by the military and industry. 


ARPA was involved in laser R&D fiom shortly after Townes' first publications on 
the laser concept in 1958. ARPA Order 6. task 12, of March 1959. provided substantial 
funding for "laser studies" in support of a broad exploration effon proposed by TRG, Inc.* 
In 196U Ted Maiman, in a Hughes Company internally-funded project, demonstrated the 
first operating laser, using a luby rod as the basing" medium. 

Soon afterwards concerns rose about the question of high energy laser beam 
weapons and the ARPA laser effort was greatiy expanded under project DEFENDER in 
order to explore its possibilities as a weapon for ballistic missiles defense (BMD).^ While 
such a development could have a very high payoff, it was considered very risky, with 
much more demanding problems than low-energy applications such as rangefinders and 
targeting devices then pursued by the Aimy.^ 

1 "Laser Pioneer Interviews." High Tech Publications, Inc., 1985. interview with Gordon Gould, p. 77. 

2 An account of the ARPA-IDA interactions leading to this expansion is given in "How the Military 
Responded to the Laser," by R. Seidel, in Physics Today, Oct 1988, p. 41. 

3 The Army and Air Force also had high power laser programs beginning at about the same time as 
ARPA Cf e g . "History of the U.S. Army Missile Command 1962-77," Historical Monograph. 
U.S. Anny Missile Command. Chapter IX. p. 169. The Navy's Office of Naval Research which did 
not have a large laser program, was used by ARPA as a main agent (AO 356 of 5/62, 9 JM.) lor Uie 
first phase of high energy laser effort Cf. also Physics Today, ibid. 


After the early exploratory work, the ARPA HEL effort was conducted in four 
phases. The first phase, lasting roughly from 1962 to 1965, encompassed a broad 
exploration of laser mechanisms, materials, and techniques for high-energy lasers.'* Ml 
prospective laser media: gases, Uquids including dyes, crystalline and amorphous solids 
were investigated. This first effort was predominandy on soUds because it appeared that 
only condensed lasing media could achieve high energy densities. The investigations 
included studies of optical and thcnnal properties and ways to improve them; damage 
mechanisms; gas flash lamps and semiconductor sources for pumping,^ "Q switching" 
rapid energy dumping techniques, pulsed power sources, and propagation of high energy 
beams through the atmosphere. Tht interaction of intense laser beams with materials begMi 
to be studied with ARPA support, at the Air Force Weapons Laboratory 

The properties of existing lasers were improved under the ARPA program, and the 
potential for high-energy appUcations of the many new lasers appearing at the time were 
investigated. Serious problems were soon uncovered, with respect to low pumping 
efficiency, thermal effects in laser generating media, and in high-energy laser beam 
propagation. An early JASON Summer Smdy indicated that the best candidate lasers, 
when scaled up to parameter ranges of interest for beam weapons, appeared to be very 
large and expensive. Further, any such beam weapon was weather-limited. It seemed 
clear by the end of this phase. 1965, that early development of a laser beam weapon was 
not likely. 

One of the most important specific technological results of this phase was the 
technique for cleaning tiny platinum inclusions from glass, which could cause explosions at 
high energy densities. This technique has also evenmaUy impacted development of all 
types of glass lasers, from low-energy systems such as range finders to medium energy 
industrial laser systems, and has been a major factor affecting the laser fusion research 
program: the high energy laser NOVA, at Lawrence Livermore Laboratory, uses glass 
technology in their Inertial Confinement Fusion program.' 

4 Robert W. Seidd. -From Glow to Row: A History of MiUtary User Research and Development in 
ffiS sSdS in the PhysUal Sciences, Vol. 18 #1. 1987. p. 111-147. and Physics Today, ibid,. 
P-36. . 

5 To use semiconductors for pumping sources was not very promising l^y^f^l ''^"^^ ""^^"^ 
a serious prospecu see Robert U Byer. "Diode Laser-Pumped Solid Slate Lasers. Science, VoL 

pp. 742.747, February 1988. 

6 As part of AO 356 of 5/62. 

7 TTic first high power glass system was apparcnUy developed in France in the late i960s. 


In about 1965 a new phase of ARPA high-energy laser effort started which 
emphasized fundamental processes and problems of scaling new lasers^ such as the CO2- 
N2 laser discovered by Patel of Bell Laboratories in 1964, to high energy. This phase of 
laser effort, however, was not as large as its predecessor.^ 

The discovery by AVCO of the high power infrared CO2 gas dynamic laser (GDL) 
in 1966 demonstrated that rapidly flowing excited gases could provide a high energy laser 
source. The AVCO laser combined two concepts. One was the work of A. Kantrowitz in 
the late 1940's on delayed equilibration in the rapid expansion of hot molecular gases 
through an aerodynamic nozzle, which suggested a way of providing an excess population 
of excited CO2 molecules. The other was the C02-N2 laser mechanism discovered by 
Patel, mentioned above. The rapid gas flow also provided a mechanism for heat 

After some delay in acceptance of the potential of the AVCO approach, in the late 
1960s another major phase of the ARPA effort toward a high energy laser began, with the 
"Eighth Card" program, under the Strategic Technology Office, classified partly because of 
the apparent potential of the gas dynamic CO2 lasers to be scaled up in energy.^ Besides 
investigation of technology and problems of the Gas Dynamic Lasers (GDL's) a number of 
new high energy gas lasers were developed with ARPA and other sources of support. 
Some of these were closed-cycle, including lasers based on flowing gases undergoing 
chemical reactions, or excited by electrical discharge or electron beams (e-beaoos), with 
improved efficiencies. ARPA emphasis in this period was on the feasibility of scaling up 
these new types of continuous wave (CW) lasers, to achieve megawatt (MW) power levels. 

8 A sampling: AO 744 of 6/3/65 called for an advanced scanning laser radar; AO 1279 of June 1968 for 
"Optical Radar." 1503 for "Ruby Laser." 2075 of March 1972 for a "Solid-State Laser Illuminator and 
2165 of March 1972 "Laser Back Scatter Studies;" 2211 of 9/72 "Advanced Lightweight Laser 
Designator and Ranger, 2560 of 8/73 for a "Multipulse Laser Target Designator." 

9 The delay is described by Seidel, Ref. 3, p. 140. A brief history is also given in pp. S33-34 of 
Reviews of Modern Physics. Vol. 59. No. 3, Part H, July 1987. A.O, 1256 "Eighth Card", 6/68. In 
the mid 1960s also, in response to Vietnam, ARPA's project AGILE looked into low energy laser 
system applications. Much of this work was under the AGILE Advanced Sensor Office and produced 
several prototype laser radars, target designators, and illumination systems which differed from those 
developed by the Army and Air Force by being lighter, smaller, and achieving new levels of 
performance. Later, a number of similar systems were developed by ARPA's Tactical Technology 

1 0 The United Technologies Research Center publication, The Researcher, October 1985. dedicated 10 the 
laser, gives a chronology of one major company's activity. 


in a reasonably sized dcvice.ii Apparently, however, the first high-energy CO2 laser of 
pulsed c-beam type was developed by Los Alamos, for their laser fusion program. 

As a result of the intense efforts in the late 1960s by ARPA and the Services, 
expectations rose that some of these high-powered infrared lasers might actually be 
engineered into a weapons system. A Defense Science Board review of the progress 
recommended in 1968, a tri-servicc laser program with each service providing its own "test 
bed" related to its characteristic platforms, with ARPA initially in an overall coordinating 

A little later, DDR&E undertook coordination of the large HEL programs, and 
ARPA's program turned more to investigation of limiting factors such as materials, optics, 
and atmospheric propagation. About this same time also several companies involved in the 
Eighth Card and other related programs began to make substantial investments in these new 
types of lasers for material processing applications. These efforts, as well as those 
supported by the military, shared many problems of optical technology, notably windows 
for high energy infrared transnoission. The damage mechanisms that had been investigated 
in a laser weapon context were important also for the industrial laser applications. A 
number of ARPA Orders from the Materials program addressed these problems, Some 
of those involved in related optical work in industry at the time have given a good statement 
of the situation: 

How much power can it lake?" **What*s tiie damage threshold?" "How 
many hours will it last?" - these were the types of questions customers 
were asking. And the answers were not readily available. New substrate 
materials to transmit high energy beams, new methods to fabricate these 
materials and new coatings able to withstand high energy densities all had to 
be developed before this situation could even begin to he remedied. 

In the late 1960s and early 1970s various government agencies realized that 
an enormous amount of work would be required to solve these problems, 
and the optical industry would not be able to handle the job without a large 
influx of funds and talent. The R&D programs thereafter established 
brought an impressive array of solid state, metallurgical, optical, and laser 

1 1 Discussion with Dr. R. Cooper, 1/90. 

12 See e.g., "High Power, Short Pulse CO2 Laser Systems for Inertial Confinement Fusion," by 
S. Singer, et al., in "Developments in High-Power Lasers and Their Applications," ed. C. Pellegrini. 
North Holland. 1981, p. 724. 

13 E.g., AO 2014 of 12/71 on Halides for High Power Laser Windows; 2138 of 2/72 on IR Laser 
Windows: 2980 on KBr for High Power IR Laser Windows, in 12/74. 

14 From "Transmission Optics for High Power CO2 Lasers; Practical Considerations" by G.H. Sherman 
and G J. Frazier. Optical Engineering Vol. 17 #3. May-Jtme 1978. p. 225. 


specialists to bear on the important problcnis, and understanding of the 
critical parameters progressed quickly. 

In this same time period, the COz laser was just beginning to establish itself 
as a viable industrial tool. The new materials processes and coatings 
developed under the various government funded R&D efforts provided the 
optical industry enough background and direction to enable it to solve many 
of the optics problems facing high power CO2 laser manufacturers and 
users. The increased laser reliability and stability resulting from unproved 
optical components helped the industrial market expand rapidly, bringing us 
to the present time, where high power CO2 lasers are being used in material 
processing applications in virtually every major industry. The hundreds of 
lasers operating thousands of hours in harsh industrial environments have 
generated a large amount of useful data and practical field experience which, 
when combined with die R&D efforts aUuded to above, finally have built a 
solid foundation of knowledge and expertise from which die optical 
industry can draw. 

Another JASON Smdy in this period indicated that practical implementation of high 
energy lasers for military use remained very difScult^^ A significant proposal to ARPA by 
Lincoln Laboratory for a large scale demonstration and test fecility, in 1969, was mmed 
down by an outside review committce.i^ A high point of this phase of DARPA effort was 
the construction in 1975, in a joint program with the Navy, of the "Mid Infrared Chemical 
Laser" incorporating the most advanced chemical laser technology achieved at that time.i^ 
MIRACL evenmally reached MW power range in continuous wave (CW) operation at near 
diffraction-limited output 1* 

Several demonstrations of lethality of the different Services* high powered gas 
lasers were also made in this period. The Airborne Laser Laboratory (ALL), initially a joint 
Air Force-DARPA effort, incorporating a United Technologies (UT) compact closed-cycle 
CO2 laser, was one of the most advanced and the longest lived of these lasers, evenmally 
achieving near-MW level power output.^^ ALL remained in R&D use until the mid 1980s. 
However, partly as a result of die JASON smdy, DARPA terminated its support of ALL in 
the mid 1970s.20 There were many discussions and proposals for laser weapons system 
applications, but apparendy none were sufBcientiy attractive to the Services. 

^ ^ Communication from Dr. E. Rechtin, 10/89. 

1 6 R. Cooper, ibid. 

17 AO 2607 of 8/73. MIRACL. 

1 8 Reviews of Modern Physics: ibid, p, S39. 

19 Ibid„p.S38. 

20 Dr. E. Rechtin, ibid. 


Problems of efficiency, size, difficulty in handling chemical systems and changes in 
operational considerations seem panly responsible. However, the MIRACX has been 
upgraded and used for several R&D projects, for the SDIO. Fig. 1 is a depiction of the 
MIRACL beam direaor. 

Figure 1. MIRACL and Navy 8EALITE Beam Director 

A major spin-off of this phase of the DARPA high energy laser effort has been to 
the industrial applications of the laser concepts and technology to materials processing 
applications as indicated in the quotation above. A more detailed perspective on industrial 
laser technology is given by some recent publications by LLNL and the National Academy 
of Engineering. The LLNL repori^^ discusses the use of Nd-doped glass in tiie NOVA 
laser used in their inertial fusion research program, and also, more generally, the status of 

21 "The Potcniial of High Average Power Solid-State Lasers." J.C. Emmcu. WJ. Knipke, and W.R, 
Sooy, LLNL Report UCRL 53571. SepL 1974. 


industrial application of medium power lasers of which much of the technology was 
stimulated by the ARPA high energy laser efforts. 

Lasers are being used for cutting, drilling, welding, and heat-treating 
operations on metals» and, as relevant on wood, plastics, ceramics, fabrics, 
rubber, semiconductors, and paper. Despite early resistance by the usually 
conservative manufacturing community, these applications have grown, and 
they constitute the largest market area for production lasers and laser 
systems. The current market is roughly split between CO2 andneodymium 
lasers, with cw CO2 lasers the only entry for applications between 400 W 
(the upper limit on neodymium) and 25 kW (the upper limit for CO2 lasers 
engineered for a manufacturing environment ). Below 400 W neodymium 
is the major entry, but CO2 competes in diat range also, and a variety of 
other laser types are reaching sufficient maturity to enter the maricet On the 
high-power side, experiments have been extended up to 100 kW but 
commercial interest is largely below 25 kW. It appears that for some time 
the advances in laser fabrication will be in the form of cost reduction, 
improved reliability, and expansion in the existing maricetplace. 

One of the most successful speciHc industrial applications seems to have been 
United Technologies Hamilton Standard laser welding system. While the power level of 
the welding laser system is considerably lower than for a weapon, the invention of this 
specific type of laser at UT (the high power forced flow, electric discharge CO2 laser) 
appears to have been definitely stimulated by the Kghth Card program, under which a high 
power version was constructed in Florida and another was used in the AFWL ALL 
program. According to Dr. AJ. Dc Maria, head of UTs laser program, the ratio of 
company funding to DARPA funding was typically three-to-one in this period, but the 
DARPA funding was always vital to maintain the company's interest to continue the 

A National Academy of Engineering publication celebrating the 25th anniversary of 
the discovery of the lasers points out that the material-working segment of the maricet for 
lasers was estimated as about $1/4 billion in 1984 with expansion expected to continue.^ 
While the direct laser market is often taken as a measure of the worth of laser technology, 
the indirect value of the laser in reducing manufacturing costs, e.g., of the industrial 
medium power laser's use in making military turbine engines, providing more efficient 

22 Discussion with Dr. De Nfaria 1/13/88. Dr. De Maria stated that the United Technology laser welding 
group is now one of iheir profit centers. 

"Lasers, Invention to Application," J.R. Whinncry et al.. National Academy of Engineering, 
Washington, D.C. 1987, p. 22. By 1983, die overall (high and low energy) laser commercial market 
was dominanL See "Lasers the First Twenty-five Years," by A J. De Maria, Optics News, Vol. 11. 
No. 10. Oct 1985. p. 87. 


machining and hole drilUng, particularly in hard and exotic materials, is probably much 

The fourth phase of DARPA high energy laser effort, beginning in the mid 1970's 
and lasting until recently, involved a return to exploration of advanced laser technology, 
along with a more directed effort toward laser systems for space applications. In the first 
part of this phase there was a strong push toward shorter wavelength, high-energy lasers, 
which could use smaller optics for the same beam quality, advantageous for space and 
other applications. Several otiicr ARPA programs in this same time period also required 
lasers in the blue green, favorable for transmission in the sea: optical communication with 
submarines, detection of submarines from aircraft, and deep underwater imaging.^s With 
ARPA (in this time frame becoming DARPA) stimulation, a number of high energy short 
wavelength lasers were developed, including, in the mid 1970*s, excimer-type and free- 
electron lasers. This effort extended to X-ray lasers, also in Uie mid 1970's.26 Much 
DARPA support in this phase went into developing other optical elements for use witii tiie 
short wavelength lasers, such as pointing and tracking controls and techniques for space 
systems, and into optical compensation techniques for the effects of atmospheric 
irregularities. An adaptive minor technique for atmospheric compensation was developed 
by Lincoln Laboratory under tiie DARPA program and has been tested using the AMOS 
(ARPA midcouise optical station) facility with SDI support 

Substantial efforts during this time period also went into developing conapaci 
efficient chemical lasers for use in space. A major product of this work was the ALPHA, a 
lightweight chemical IR laser system. The ARPA space laser system program, including 
ALPHA, large space optics, and pointing and tracking in space^^ eventually became tiie 
TRL\D program. This technology also was transferred to the SDI effort 

One of Uie main efforts under tiie SDI program to explore tiie potential of tunable 
high power free electron (FEL) lasers has used a induction accelerator generating a 
relativistic, high intensity electron beam, tiie Lawrence Livermore Laboratory's advanced 

24 ' Dr. AJ. DeMaria, UT Inc., discussion Januaiy 1988. 

25 c.g.. AO 1871, of 5/71 and 3588 of March 1978. 

26 e.g.. AO 2694 of January 1974. The first successful X-iay lasers, however, apparcndy occuned in the 
eariy 80's, under ihe laser fusion program ai the Lawrence Uvermorc Laboratoiy. 

27 AO's 2761 of 7/74, 3526 of December 1977 and 3945 of Febniary 1980. ALPHA is briefly described 
in Reviews of Modem Physics, p. 539. 


test accelerator (ATA).^^ The ATA accelerator was not funded by the ARPA Laser 
program but by a different ARPA effon which was aimed at exploration of the potential of 
particle beams for directed energy weapons. The ARPA particle beam program had its 
origins in 1958, and disappeared in the late 1960s, but came back in the mid 1970s and is 
still part of the DARPA long range Directed Energy Weapons Program. 

Other potential, smaller space-based laser system applications, such as for air 
defense, were also investigated by DARPA in this period 2^ In the late 1970's. DARPA 
commenced a joint program with the Navy toward a blue-green laser system for 
communicating with submarines. Initially, this program was closely related to the short 
wavelength, high energy laser prognun. It included two approaches: a ground-based 
laser-satellite mirror combination, and a space-based laser. The ground-based laser system 
and adaptive mirror combination was tested at AMOS as mentioned above. This program 
was transferred to SDI. The space based laser approach continued and, after DARPA 
development and demonstration of a suitable narrowband filter optical receiver and a 
matched wavelength laser, (a modest energy ultraviolet excimer laser product of the 
DARPA short-wavelength effort) pumping a lead vapor "Raman" converter cell) and 
commencement of effort toward making the laser system qualified for space, this program, 
now named SLCSAT. was transferred to the Navy.^^ However, a recent Navy-D ARPA 
MOU addresses continuing investigation of solid state lasers considered more suitable for 
space than the gas excimer lasers. 

A significant spin-off of the DARPA short wavelength laser effort was the copper 
vapor laser. This laser was actually invented during the early TRG effort in the mid 
1960's, and was further developed at GE in the late 1970's witii support from the DARPA 
short wavelength laser program. The copper vapor laser is now a commercial product, and 
is the pumping laser for tunable dye lasers in the DOE's Livermore Laboratory atomic 
vapor laser isotope separation system (AVLIS), which was the preferred approach for the 
DoE nuclear fuels enrichment program.^^ 

28 Very recently, however, SDIO has selected a different approach to ihe free-electron laser, based on a 
radio frequency driven accelerator. Cf., Aviation Week, Oct 23, 1989, p. 21. 

29 R. Cooper, ibid. 

"Submarine Laser Communication," by Comdr. Ralph Chatham, J. of Electronic Defense, March 
1987. p. 63. 

^ ^ See Laser Technology-Development and Applications, Hearings before the SubcommiiKe on Science, 
Technology and Space of ihe Committee on Science, Technology and Space, U.S. Senate, 96th 
Congress, December 1979, p. 78-79; also, DoE Annual Report to Congress. 1986, p. 151. The 
Copper Vapor Laser was invented by Gould at TRG in 1966. see "Efficient Pulsed Gas Discharge 



ARPA's initial involvement with lasers was through an unsolicited proposal from a 
pioneering industrial group. This effort, however, did not yield any breakthroughs. After 
the first working laser was developed elsewhere, there were speculations in ARPA and 
OA that this new area could have very high military potential and ARPA soon set up a 
sizeable effort in the high payoff and very risky high energy lasers area for weapons. 
Since this ARPA program began so close to the time of origin of a new idea in physics, it 
was a complex high technology effort with many players to more confidcntiy dctennine and 
assess the payoffs, the limiting factors, and. importantly, the potential threat The Army 
and the Air Force also had large laser programs, at about the same time, and the AEC 
developed a large high energy laser program for the incrtial confinement fusion QCP) 

Some feel this early ARPA effort should have been curtailed earlier than it was. An 
early JASON assessment pointed out limitations due to propagation and the size of any 
prospective weapon system using the available technology. However, Uiere were many 
uncertainties in propagation efficiency, pointing and tracking, letiiaUty, and practicality of 
weapon systems. Many different kinds of lasers were being discovered-almostaU outside 
the military programs. All this and the high potential payoff made such a program decision 
difficult. ARPA also had some of the best available advice for its early actions.32 xhe 
reason for continuing a high level of ARPA effort at this time may have been that some felt 
that better glass cleanup might overcome the problems.33 fact, the glass laser 
technology developed in this phase under ARPA support has had a major impact on almost 

Lasers," by W.T. Walter. N. SoUmcne/M. Piltch, and G. Gould, IEEE Journal of Quantum 
Electronics, V. QE-2. Sept 1966, p. 474-479. but significant further development was necessary W 
become practically useful. According to Dr. T. Karras of G£., much of this development was funded 
by DARPA Considerable further development for AVUS occurred at Livcrmore. Discussion wiih 
T. Karras. A.O. 3650 of 7/78. Very reccnUy, however, DoE has ordered a new review of aU 
enrichment technologies, and has apparently put off further AVUS development 

32 c. Townes. the inventor of the laser, was at IDA during this period. Apparentiy, however, Townes did 
not seem to be a strong advocate of the high energy laser program. Discussion widi Dr. C. Cook, 

33 Discussion with R. Collins, IDA, June 1989. 


all subsequent laser work involving glass. However, at one stage the French had produced 
the best glass, which was purchased by the U.S. programs. A number of key ideas over 
the years also came from the intensive Soviet efforts. 

The invention of the gas dynamic laser, also from outside the ARPA program, was 
a surprise. The ideas involved were quite different from those of the previous program 
which emphasized solid laser media. There seemed good reason to "step on the gas" 
because the GDL technology appeared to be scalable to high energies. The large "Eighth 
Card" ARPA program, along with service and ICF programs, provided the climate for 
rapid developments of several derivative types of infrared lasers. Window and minor 
materials were soon indicated as limiting factors. The ARPA materials program gave 
essential help to solve many of these problems, and ARPA's efforts to disseminate 
information on laser damage of optical materials was of great value to industry .3* The 
three services became heavily involved. ARPA, besides supporting advanced technology 
and investigating limiting factors of possible systems, was given a coordinating role, which 
was later taken over by DDR&E and the DoD HELRG (High Energy Laser Review 
Group). Joining with the Navy, ARPA produced at the end of the 1970's a high power 
laser system, the MIRACL, which is still regarded as close to the state of the art, has been 
upgraded for use in SDI R&D, and may be again for AS AT application-^^ 

Some feel that this expensive period of system oriented development could have 
been avoided if there had been agreement, in the late 1970s to prosecute a well coordinated 
program in a simple major facility .^^ Others point out that, during this period, because of 
the program's classification, contacts with the "outside" laser community, which were 
carrying on substantial efforts, were largely cut off, and that had it been possible to 
maintain these contacts, a more realistic program may have been pursued,^'' In fact, some 
contact was maintained through the HELRG. However, the impact of this phase of 
ARPA's effort on industrial use of moderately high energy gas lasers has been substantial. 

ARPA was rather "responsive" to outside developments in the first phases. 
However, when the long wavelengtii technology had matured enough to make more 
realistic estimates of what would be required for weapons systems, DARPA began to 
support more directed work toward the objective of shorter wavelength lasers. This 

See e.g., Taser Induced Damage to Optical Mirrors," National Bureau of Standards, Dec. 1976. 
35 Aviation Week, December 19, 1988, p. 29. 
3^ R. Cooper, ibid, 

R. Collins, ibid. 


DARPA program helped develop several types of new shon wavelength lasers, in the 
visible and ultraviolet, one of which, the "free electron laser" (invented sometime earlier 
outside the DARPA program), profited from the availabUity of the ASTRON accelerator 
facility at Uvcnnore, partly developed under the separate ARPA particle beam weapon 
program. X-ray lasers were investigated under this program but abandoned a few years 
before success was reported by the Livermore ICF laser group. Some DARPA suppon 
was apparently given to the bomb-driven X-ray laser work at Uvcnnore, before SDIO was 

A joint program with the Navy for submarine laser communications profited greatly 
from cxcimer laser work, carried out under the DARPA short wavelength laser effort, and 
has led to demonstration of a workable, moderate power, laser-optical receiver 
combination. Recently, however, the Navy and DARPA have agreed that the risks and 
expenses in developing new solid state lasers for the blue-green, are perfiaps more 
acceptable than those associated with going ahead with the gas cxcimer laser systems in 
space. The motif for conomunication needs also benefited the DARPA laser effort in 
providing a motivation which allowed atmospheric compensation experiments, relevant to 
the laser weapons program, to be canied out at more convenient lower laser powers. 

The SDI has depended heavily on the DARPA laser technology, notably for the 
MIRACL, ALPHA, and the associated TRIAD pointing and tracking systems, and the 

The overall military high energy laser effort has been criticized generally as being 
overly ambitious and costly, with no resulting system in the inventory. Another criticism 
has been that limiting factors were soon discovered, which should have discouraged 
attempts to develop high energy laser weapon systems. Perhaps the problem of a "closed" 
conomunity in which, because of the newness of the field, the contractors have a more 
deterministic role, led to excessive efforts. However, because of the wide "public" 
appreciation of the very high potential payoff, related concerns about potential threats, and 
die high unit cost of a R&D item in this field, it is difficult to see how DARPA could have 
done very differently. DARPA's role was to develop the new technology, and to construct 
state-of-the-art devices. Without a solid knowledge of the technology and its limiting 
factors, and of die practical difficulties in the construction and operations of high-energy 

38 "Excalibur." A.O. #4557. 4/82, for $7.9 million. 


lasers, it would have been very difficult to make a good assessment of potential threats in 
this area. 

On the positive side, due to the DARPA program, state-of-the-art high energy lasers 
have been produced, and arc being used by military R&D programs. There have been 
substantial spin-offs to lower energy military systems and to industry and the fact that the 
military R&D facilities and many of the spin-offs exist at this time, together with a strong 
technological community, can be largely credited to the DARPA program. 

DARPA's total investment in lasers has been the largest in the military, estimated 
from project records as about $3/4 billion.39 The direct value of the material-working 
medium power industrial laser market has been estimated as close to $1/2 billion, DoE 
expenditures for Copper Vapor Lasers in the development of the AVLIS technique are 
estimated at about $3/4 billion.^ 

Counting in the space mirrors work this approaches $1 billion. 
^0 Lawrence Livcrmore National Laboratory, Institutional Plan FY 1985/90, pp. 11 8- 19. 












The ARPA (and DARPA) involvement in Over the Horizon (OTH) high-ftequency 
radar between 1958 and 1975 can be described as a successful effort in coordination, 
exploration and development of technology. One of the first payoffs was technology in the 
early 1960's for what became the Air Force 440L early warning system, which was 
deployed in 1966 and retired in 1975 when satellite systems for early warning became 
operational. Another spin-off was an oblique chirpsounder now in use in tiie AN/TRQ-35 
frequency selection system for high-frequency military radio communications. DARPA- 
developed OTH technology had a major impact on the Air Force FPS-1 18 OTH-B radar 
system for CONUS air defense, approaching full operational deployment.^ and on the 
Navy OTH-R system for air defense now in full-scale development.^ DARPA OTH 
technology also provided much of the basis for the Australian OTH System for tiiat 
nation's air defense.^ 


Electromagnetic waves in die high-frequency band (with wavelengths of tens of 
meters) reflect downward when incident obliquely on ionospheric layers at hundreds of 
kilometers altitude. In this way electromagnetic energy can be propagated in a "guide" 
between eartii and ionosphere to thousands of km range, a phenomenon long in use in 
high-frequency radio communications. This concept forms the basis for OTH radar. 

The history of OTH radar apparently goes back at least to WW n, when an 
experiment during the development of the British CH (Chain Home) Radar Air Warning 
System, which operated in the upper end of the high frequency band, large diffuse echoes 
were observed and attributed to backscaner from the earth, after ionospheric reflection, at 

1 "Backscatter Radar Extends Warning Times." David A. Boutacoff, Defense Electronics, May 
p. 71-83. 

2 "The Frontier of Sensor Technology." by LCDR J, Sylden, USN. Signal, March 1987. p. 73-76. 

3 The Defense of Australia, Australian Deparunent of Defense. 1987. p. 4 and p. 35, 


ranges up to several thousand miles.-* Shortly after WW H there were smdies and some Air 
Force-supported experiments in the US, to detect aircraft and V-2 missUcs* without much 

When ARPA began in 1958 there were several active mUitary efforts under way. 
At the Naval Research Laboratory work had been going on since the early 1950's using a 
pulse-doppler radar with a great deal of signal processing to remove the large earth 
backscaner background for low-altimde targets and related propagation studies.^ The 
"MUSIC" NRL effort was supported by the Air Force as an approach to long-range 
detection of aircraft, up until 1958 the highest priority. Another OTH effort had been 
conducted for some time by the Air Force's Cambridge Research Laboratory (AFCRL). A 
third, under project "Tepee" sponsored by ONR, had a later stan in 1956, exploring 
initially the possibilities of using available equipment of the type then used in CQP 
(Communication Zone Indicator) smdies during the IGY to detect, first, nuclear explosions 
and, later, ballistic missiles, both of which might have large radar cross sections and/or 
cause large ionospheric disturbances. Some of this ONR-supported work was done by a 
Stanford group under O.G. ViUard, which had been conducting ionospheric studies with 
other ONR electronics research support for some time. 

Because of the high priority of ballistic missile defense and ARPA's broad 
responsibilities and funds under project DEFENDER, OTH R&D began to be coordinated 
under ARPA.^ ARPA also began to support exploratory, high-risk R&D on a wide range 
of OTH techniques and problems, such as antennas and receivers, ionospheric 
propagation, signal formats, management of interference, and ionospheric sounders.? 
Much of the research was done by the Stanford Group, which also served as advisors for 
the ARPA program. 

4 -Radar Days," by E.G. Bowcn. Hilgcr 1987. pp. 13-14. ApparenUy theie was an idenuficauon of 
ground, backscattcr echoes, called "Splash backs." in pulsed round the «'?5W propagation expenmen^^^^^^ 
NRL in 1926. See "Evolution of Naval Radio & Electronics and contnbuuons of the Naval Research 
Laboratory" by LJL Gebhard, NRL Report 8300. 1979. pp. 45. 

5 -Over the Hwizon Backscatter Radar." J.M. Headrick and I. Skolnik, Proc IEEE, June 1974. p. 664. 
Remarkable analog processing techniques were developed in the eariy NRL program. 

^ Earlier OTH coordination meetings had been conducted by ONR. 

7 E.g.. AO #32 of 10/14/58 provided nearly $3.5 milUon to ONR for OTH radar measuicmcnis. 


Many of the subsequent payoffs are traceable to this early ARP A- sponsored 
exploratory work, which extended through the early 1960's.8 One of the earliest of these 
payoffs was the work by the Stanford research group, separately supported by ARPA, on 
an approach to long range ICBM raid detection.' These efforts formed much of the basis 
of the AF 440L "forward scatter" system, which began to be operational in the late 1960's, 
at a critical time when, because of the failure of the AF Midas satellite program, diere was a 
need for an eariy warning system for detection of a massive missile attack. This relatively 
simple (and low cost) "forward scatter" system consisted of a set of transmitters in the Far 
East continually monitored by a set of receivers distributed in Europe. The main technical 
question regarding the 440L was the ionospheric stability and continuity over the 
propagation paths. Early field measurements, which incidentally detected some ballistic 
missile launches, showed that the stability was sufficient for a useful system and developed 
critical data on false alarms and failure to alarm. The 440L was retired in 1975, after 
infrared satellite early warning systems were deployed.^*^ 

Another early result from this same group was the Barry high-frequency sounder, 
using a low power, continuous-wave, digitally controlled, highly linear frequency-swept 
signal, (FM-CW). A significant achievement of this digital sweep, due to G, Barry, was 
that it preserved phase coherence, This technique and the associated digital-processor 
and receiver equipment was used to obtain high range resolution and select favorable 
frequencies for OTH radar. Later it became a key part of the AN/TRQ-35V tactical 
frequency management system for HF military communications.^^ Later experiments by 

8 Some examples of ARPA projects in this period include: AO # 90, of 5/2/60, for an OTH data 
coUecaon and analysis center at SRI; AO # 160 for $1.6M to NRL for "Music Madrc Radar Program," 
including modification of doppler processing to detect accelerating rockets and exploration of long range 
ducted propagation; AO #196 of 1/61 to explore the potential of longer range multihop HF backscatten 
AO # 299 of 1/1 1/62 exploring "Sky Waves." 

5 AFCRL had similar ideas, and was conducting experiments under project CAME BRIDGE, but 
Dr. Fubini of DoD was more impressed with the Stanford approach and data, and jHescribed that it be 
used. AFCRL news release 5/68 and discussion with Dr. Villard» 7/88. 
"History of Strategic Defense/ by C.W. Mausu ei al., SPC Report 742, 1981, p. 3. 

1 1 The digital sweep generator was originally suggested by Villard when the Hewleu-Packard digital 
frequency synthesizer became available. The modification to a coherent synthesizer by Barry was later 
adopted by Hewlett Packard. Communication from O.C. Villard 1/90. 

^ 2 Acceptance of the Baity Sounder, which became a commercial product in die 1960's, was based on AF 
trials in the early 1970's. Cf. "Real Time Adaptive Frequency Management," by Robert B. Fenwick 
and Gerard J. Woodhouse, in "Special Topics in HF Propagation," ed. V J. Caycc, NATO AGARD 
Congress Proceedings, # 263, 1979, pp. 5-1 to 5-14. Earlier Navy poor experience with a major 
invesunent in other HF sounders led to rejection of the Barry Sounder for nearly 10 years. Discussion 
with Dr. G. Barry 4/5/88. 


the Stanford group demonstrated the advantages of this digital- linear FM-CW signal foraiat 
for OTH backscatter radars, and the same signal format is now used in the OTH backscattcr 
systems being deployed by the Air Force, Navy, and Australians. 

During diis same period, the NRL OTH group continued work on the MUSIC- 
MADRE experimental OTH pulse-dopplcr radar. In 1960, ARPA funds provided for 
modification of NRL doppler processing to improve detection of high acceleration missile 
targets, and for development of other techniques. ARPA support was veiy important to the 
NRL project because the air defense motif for die NRL work waned in the late 1950 s and 
eariy 1960's due to the priority attention then being given to baUistic missile defense.i3 The 
long-range air-defense motif returned strongly, however, in the late 1960's, This motif 
was largely responsible for the fact that OTH remained in ARPA when DEFENDER was 
transferred to the Army in 1967. 

In 1963 the Air Force proposed and OSD accepted, in principle, a future Ah- 
Defense modernization program, including AWACS and OTH backscatter radars.i^ in 
1967 also, a DoD DS ARC decision affirmed CONUS air defense as an objective for OTH. 

In die mid 1960's to early 1970's, performance limits of wide aperture non-rigid 
HF antenna technology were tested by the Stanford group with ARPA support. The NRL- 
OTH radar, which made most of the earliest backscatter detections, used a rigid antenna to 
avoid spurious doppler effects during long integration times. It was not certain how much 
could be done with wider but less rigid antennas. The Stanford Wide Aperture Research 
FaciUty (WARF), with a 2.5 km aperture, much wider than any before attempted (see Fig. 
1) was constructed in 1966. mainly witii ARPA support 

The WARF width was determined after a number of experiments, together witii 
practical enginering considerations.i5 InitiaUy. the low-powered WARF was not expected 
to detect aircraft^s However, high resolution in azimuth and range was found possible 
using the WARF. which, togetiier with sophisticated digital processing of the highly linear 
digital FM-CW signal, allowed detection and tracking of aircraft and the systematic study 
of this capabiUty as functions of radar parameters. The WARF experiments estabUshed 

13 A 0 160 of 6/60 to NRL for Music Madre. The additional support is credited with getting the 
^^^l^comp^ttd in Gebhaid. ibid., p. 126. Sec also "History of Strategic Defense, ibid., 

p. 9. 

14 Communication from Dr. A. Flax. IDA, 2/90. 

1 5 Support of WARF was also given by ONR- 

1 6 Discussion with Dr. L. Sweeney of SRI, 4/6/88. 


many benchmarks for performance for later systems, and also laid the basis for automatic 
detection and tracking techniques. This technology was transferred effectively, and 
infonnally, in the regular OTH symposia run by the ARPA program director. In particular, 
the Air Force adopted the FM-CW signal format and separate transmitter and receiving 
antennas for its future OTH radars in the early 1960s, for their 441B and 118L systems. 

In 1967 ARPA began to plan project BIG PUSH, aimed at an experimental system 
embodying the state of the an of pulse doppler and FM-CW technology, with flexible 
characteristics enabling detection and tracking of a variety of targets, including baUistic 
missiles at long ranges, and aircraft. BIG PUSH incorporated high. 

Figure 1. WARF System 

power and a variety of waveforms, the highest aperture achievable and up to date digital 
processing. However, BIG PUSH was not approved by DoD, on the grounds that the Air 
Force's large FPS-95 radar project was dien under way, and DoD could not have two large 
competitive OTH research projects at the same time. The FPS.95 was a high power pulse 
doppicr system with a unique antenna, and was turned off after a short period of 
unsuccessful operations. The FPS-95 experience had quite a negative impact for some time 
on much DoD thinking about the eventual utility of OTRi^ aRPA, however, continued its 
OTH program, albeit somewhat reduced, despite the unfavorable climate. 

In the early 1970's WARF experiments also examined the potential of OTH for sea 
state and wind patterns determination. This led to demonstrations in die late 1970's of die 
WARFs abUity to remotely track hurricanes in the Gulf of Mexico.i^ Later, taking 
advantage of HF propagation management possible with new processing capabiUties to 
isolate single propagation modes, ships were detected using die WARF.^' 

The ARPA program turned, in die early 1970's, to the problem of evaluating risks 
for OTH for detection of aircraft in die higher latitudes, widi the singular auroral and polar 
cap ionospheres. A strong motive for this investigation was die fact diat CONUS air 
defense would have to deal with diis northern section. A number of experiments were 
performed, and analyzed under the joint ARPA-Air Force "Polar Fox"20 experiments, 
which explored the capabilities of OTH backscatter radars, both pulse-doppler and FM- 
CW, in die mid to higher latitudes, and auroral ionospheric regions marked by spurious 
reflection and propagation. A somewhat later project, "Polar Cap," explored these 
capabilities in the polar ionospheric region, within the Auroral ring, marked by 
irregularities and absorption. The results of these experimental projects were used for die 
assessment of die statistical probability of detection in diese regions by OTH systems, 
which because of die large scale coverage would have many opportunities during a large air 
attack. The results affected die later decisions on siting and orientation of CONUS OTH air 

1'^ Discussion with Dr. C. Cook, ex-ASD for Defensive Systems, 12/89. 

18 "High Frequency Sky Wave Radar Measurements of Humcane Anita," by Joseph W. Maresca and 
Christopher T. Carlson, Science. Vol. 209, 12 Sept, 1980, p. 1189. 

19 "Ships Detection With HF Sky Wave Radar," J.R. Bamum. (IEEE) ^cea't EngiMering.W^^^^^ 

No 2 AorU 1986 Large ship detections were first demonstrated by NRL in 1967, See Kci. 4. ine 
ARPA siippoit to NRL was key to development of a digital filter that was used for these detecuons^ 
Discussion with J. Headrick, NRL 6^8. During WW U. U.K. re^hers apparenUy considered OTH 
radar for detecting convoys. Communications from O. Villard. 1/90. 

20 E.g.. AO 1765, Of im. 


defense radars generally away from the auroral regions.^^ Data from these northern 
experiments were also valuable for assessment of effects of high altitude nuclear explosions 
on military HF systems for communications and OTH. Increasing appreciation of the air 
threat to CONUS provided motivation for the Air Force finally going ahead with OTH 
backscattcr systems for CONUS defense in 1975.22 

DARPA formally transferred their OTH program to the Air Force in 1975. After its 
FPS-95 experience mentioned above, the Air Force decided to adopt the DARPA-gencratcd 
FM-CW signal format with high average power and large bandwidth together with a wide 
aperture for their OTH backscattcr radars, '^ith General Electric as contractor, RADC built 
and operated a demonstration model OTH radar in the early 1970s, which detected and 
traced aircraft at long ranges over air and watcr.^^ In 1975 the Air Force awarded a 
contract to General Electric for construction of an experimental OTH radar which was a 
prototype for continental air defense. Tests with this OTH radar were successfully 
completed in 1981. Since then several sections of the Air Force CONUS OTH FPS-118 
systems have been constructed and are approaching operational stams.2* Figure 2 shows 
one of the hardened FPS-1 18 prototype transmitter antenna fields. 

In the early 1970s, because of growing appreciation of the BACKFIRE threat, the 
Navy began to be interested in long-range detection for fleet air defense. Later a number of 
Navy Integrated Tactical Surveillance System (ITSS) studies were conducted which 
indicated that satellite capabilities for this purpose were not likely to be available before the 
1990*s, but that OTH B backscattcr radar technology, deployed to forward areas, might 
satisfy tiie need until tiien. In the late 1970s, after demonstration of ship detection, the 
Navy interest increased, and DARPA technology, especially in antenna systems, signal 
format and signal processing, played a major role in die design of the Navy relocatable 
ROTH-R system now in full-scale development. Figure 3 shows an ROTHR transmitting 
antenna field, similar to that of the WARF. 

2 1 However, the Air Force now plans to deploy an OTH backscattcr radar in Alabama to cover the "North 
Stope" BACKFIRE attack corridor. 

22 Discussion with I^. C. W. Cook, cx-ASD for Defensive Systems, 2/89. 

23 Communicauon from Gen. J. Toomay, 1/90. 

24 See Ref. 1, and also "Warning and Assessment Sensors," by MG. John C. Toomay, USAF (Ret) 
p. 292, in Managing Nuclear Operations, by Ashton B. Carter, el al., Brookings, 1987. 


Figure 3. Relocatable Over-The-Horlzon Radar (ROTH-R) Transmitting Antenna 
Field (From Director, OTE, Report to Congress, FY 1987) 

Australia had a smaU OTH program dating from the late 1950s. Early experiments 
using bistatic HF CW radar systems took place in the joint ARPA- Australian ballistic 
missile experiments in the early 1960's at the Woomera test range. As a result of an 
initiative by the Australians, a specific U.S.-Australian cooperative program in OTH began 
in the early 1970's,25 and DARPA estabUshed an office in Ausu^ia to facilitate the transfer 
of OTH technology to that nation's JINDALEE experimental OTH radar. Construction of 
the Australian operational OTH system based on JINDALEE is planned for Spring 1990.26 

25 "The Development of Ovcr-the-Horizon Radar in Australia." by D.R Sinnott, Australian Government 
Publishing Service, 1988. 

26 See Ref 3 and "The JINDALEE Over-the- Horizon Radar System," by R.H, Sinnott, paper at the 
conference on Air Power in the Defense of Australia, 14-18 July 1986, Austr^ian Nauonal Umver,^^^^^ 
See also Avia/wn Week, May 11, 1987. JINDALEE means "Bare Bones m Abongine. which Sinnou 
says characterizes the effon. 


There were also unsuccessful attempts by ARPA to explore use of other parts of the 
electromagnetic spectrum for OTH purposes, including the VLF and VHF range. 
Ionospheric modification by high-power HF transmitters was also tried in the attempt to 
generate or modify reflecting ionospheric conditions. 

OTH technology, while now considered mature, is still undergoing some 
development, paced again by advances in data processing and networking technology, and 
by in^iovements in understanding of the complexities of the ionosphere. 


DARPA's OTH program began as an approach to early warning of missile attack 
under project DEFENDER. It was built on earUcr Service programs. While it began under 
DEFENDER, it did not receive as much attention as die terminal defense DEFENDER 
programs. Like HF communications. OTH was widely regarded as partly unreliable, 
particularly in the event of nuclear exchanges, which were a major consideration in 
DEFENDER. However, it seems to have been one of only two DEFENDER programs diat 
led directly to a deployed system for warning of baUistic missile attack, in this case the 

Sustained support of a very strong Stanford (later SRI) Group under Villaid proved 
highly productive. Timely ARPA support was provided for the 440L and related 
developments in a period of crisis for ICBM attack warning. Later ARPA provided 
continuous backing through a long period of OTH technology development for air defense, 
which remmed to high priority in the late 1960's. Out of this sustained effort came two of 
the key technologies used today, although these were considered risky for many years.28 
The first of these were digital Unear frequency sweeping to generate a coherent frequency 
modulated-continuous waveform (FM-CW), (applied also with some delay, in the TRQ- 
35V system). Secondly, die program demonstrated the utiUty of high resolution obtained 
by very wide aperture, less than rigid antenna systems. This demonstration took many 
years, which was necessary to get statistical information on propagation stability. Not only 
the fr«iuency sweeping, but all the processing technology in OTH was greatiy assisted by 
the general advances in digital processing technology which occurred during the same time 
period, and were quickly applied to OTH by Stanford and die other ARPA contractors. 

27 The other was ESAR, which led direcUy to the Air Fbrce FPS-85. still used partly for SLBM warning. 

28 Communication from T. Croft, 1/90. 


The productivity of the Stanford (now SRI) group is attributed by them largely due 
to ARPA's continuous long-term suppon and "light handed" managements^ 

The ARPA BIG PUSH OTH program was an attempt to construct a state of the an 
research system. Apparently, part of the motif was to test the relative performance of FM- 
CW versus pulse doppler technology. It was stopped by DoD because of the large Air 
Force (pulsed) FPS-95 OTH radar program then under way. The FPS-95 was a result of a 
"parallel" RADC OTH program, which was recognized as a dangerous competitor, but 
apparenUy not strongly opposed by ARPA 30 Because of BIG PUSH'S cancellation the 
ARPA program transferred key technologies, and not a system. 

The long series of ARPA's OTH coordination meetings led to an effective, if 
informal, transfer of these technologies to the Air Force and later to the Navy. There were 
always some elements of competition in the DARPA OTH program, between pulse doppler 
(NRL, Indusoy) and the FM-CW techniques assessed by the Stanford group. Evenmally 
the Stanford combination of FM-CW waveform and wide aperture was agreed on by the 
community involved as the preferred approach. The unsuccessful experience with the FPS- 
95, a pulse doppler system, was crucial to the final decision by the Air Force to adopt the 
FM-CW waveform approach, ARPA's POLAR experiments provided opportunities to 
demonstrate the capabUities of OTH technology, bodi pulse doppler and FM-CW. and 
provided and key ionospheric information for Air Force decisions on OTH for CONUS air 
defense in the early 1970's. 

The Stanford-ARPA WARF technology, while not itself a prototype for the Navy's 
ROTHR systems, provided most of the essential technology for that system. The Navy's 
interest in long range air defense was in reaction to the BACKFIRE threat, and its decision 
to go ahead with ROTHR came only after its extensive ITSS studies indicated that adequate 
satellite systems would not be available until nearly the end of the cenniry. 

Increased appreciation of threats to CONUS from aircraft which could launch cruise 
missUes provided an additional challenge to this technology. The OTH air defense 
technology appears to be meeting a timely need, at least until satellite systems such as 
TEAL RUBY also largely developed with other DARPA-support, can be tested and 
deployed- The Air Force estimates its 1 1 8L system to be useful for more tiian 25 years. 

29 Discussion with L. Sweeney and T Croft. 5/88 and O.G. ViUard of SRI, on 7/88. 

30 Discussion with J. Kane and E. Lyon, 1/90. ARPA's Navy agent, however, did express opposition id 
the FPS-95. 


In retrospect, the dedication and management skiU of a single ARPA (and DARPA) 
OTH program manager, Alvin van Every, throughout the 1958-1975 period, can be 
credited for much of the program's success.^i 

DARPA-dcveloped technology formed the basis for the Australian air defense 
systemTfaciUtatcd by van Every's going there personally as DARPA's representative in 
1975. Some experts feel the Australian system has profited from more recent data on 
performance of the U.S/ OTH radars, and may be a more advanced system when built 

The total ARPA expenditures for OTH appear to have been about $100 miUion. The 
Air Force 1 18L east and west coast systems cost exceeds $1 billion, and the R01HR cost 
is estimated as more than $1 billion dollars 32 The fact that the ARPA programs had a large 
academic component, which was low cost, and that there was a single ARPA manager 
throughout, may have had an impact on the scale of the expenditures. Not everything tried 
in the D ARPA-OTH program worked, but "poor horses" were generally soon abandoned. 

The Soviets have pubUshed two books on OTH technology, the latest of which has 
been transcribed in the U.S. and refers extensively to results of U.S. OTH research.33 The 
Soviets large "WOODPECKER" OTH radar system, however, apparcnUy does not use 
FM-CW signal modulation technology, and causes much interference in the HF radio 

3 1 Van Every had also been a graduate student under ViUard. 

32 HASC DoD Appropriations Hearings. 99th Congress, 2nd Session, Part 3, 1987, p. 

33 Over the Horizon Radar, by A.A. Kolosov. ci al., Artech House. 1987. 

34 Short Wave Listening With the Experts, by Gary L. Dexter, H. Sams Co., 1986, p. 




























7-28^1 M 




AMOS (ARPA Midcourse Optical Station) was initiated by ARPA in 1961 as an 
astronomical-quality observatory to obtain precise measurements and images of reentry 
bodies and decoys, satellites and other space objects in the infrared and optical spectrum. 
Located at nearly lO.OOO-ft altimde atop Ml Haleakala, Maui, Hawaii, AMOS served as a 
unique facility for operational measurements and R&D from the early 1960's. AMOS' 
twin infrared telescopes were transfeired to Air Force in the late 1970's as MOTIF: the 
Maui Optical Tracking and Identification Facility, now regarded as one of the primary 
sensors of the Air Force Space Tracidng System, Transfer of the optical telescope and the 
remainder of a highly automated AMOS to the Air Force took place in 1984. 


The concept of AMOS was originally proposed in 1961 by R. Zirkind of the ARPA 
staff as an astronomical-quality facility for imaging reentry bodies and other space objects 
in the infrared, and for performing research in infrared astronomy. Information on the 
infrared emissions from reentry bodies in midcourse, expensive to obtain in space, was 
needed particularly for assessment of detection and discrimination systems then under 
study in the BAMBI and PRESS projects under ARPA's DEFENDER program. The 
location selected for AMOS, at about 10,000 ft altimde near the top of Ml Haleakala, the 
largest dormant volcano crater in the worid, was above most clouds and most of the 
infrared-absorbing water vapor in the atmosphere. The site was also expected to have very 
good astronomical "seeing." For similar reasons the site had been selected previously for 
one of the Baker-Nunn Satellite Cameras used to track satellites during the IGY.^ The 
AMOS location was favorable for observation of reentry vehicles and decoys, missile 
bodies and other objects over a considerable portion of the midcourse range of sub-orbital 
trajectories between the Vandenberg missile launch site and the main reentry location at 

1 "Trackers of the Skies," by E. Nelson Hayes, Howard Doyle, Cambridge 1968. p. 33-34 The 
University of Hawaii operated the Baker-Nunn telescope for the Smithsonian Astrophysical 


Kwajalein. The low-latitude location was also advantageous for observations of satellites. 
AMOS was conceived initially to include two high quaUty telescopes, one for use in the 
inftared and the other in the visible spectral region, with precision mechanical mounts and 
computer-controlled drives. 

Zirkind had a strong desire also to exploit, part-time, the capabilities of such a 
system to open a new field of astronomical research in die infifared.2 Dr. j. Ruina, ARPA 
director at the time, gave his approval to the project, provided the astronomical community 
agreed it was a good idea, and would actually do research witii AMOS. A meeting of 
several prominent astronomers was held at Harvard's Smithsonian Astrophysical 
Observatory in Summer 1961. at which it was agreed Uiat AMOS' plaiined infrared 
observing capabiUties and its location further soutii tiian tiien existing U.S. observatories, 
were indeed of interest in astronomy. ITie conclusions of this meeting, and the results of a 
careful investigation of astronomical "seeing" a Uttie later by one of the participating 
astronomers (G. Kuiper). which indicated tiiat resolution of die order of 0.1 seconds of arc 
was often attained, led to further plans for an additional, somewhat larger telescope at 
AMOS for use in the optical spectrum. 

The AMOS effort formally began with Amendment No. 2 to an existing ARPA 
Order 236, to the University of Michigan's Instimte for Science and Technology, for 
telescope design, construction, and eventual operation of the observatory .3 The ARPA 
Older amendment stated the AMOS objectives as: (1) "Identification and signature of space 
objects; (2) an active program to advance the state of the art of infrared technology and 
high-resolution imagery; (3) a research program in geophysics and astrophysics including 
die astronomical community." The Department of Astronomy of die university was 
involved in the initial design studies for AMOS. The previously mentioned "seeing" 
investigation was one of the first subcontracts, and was faciUtated by die existence of die 
existing IGY-Smithsonian Baker-Nunn telescope at the site. The AMOS site was leased 
from the University of Hawaii. The original terms of the lease provided for operation^of 
die AMOS Observatory faciUty by die University of Michigan, and after 10 years use wh«i 

"Project AMOS: An Infrared Observatory." by R. Zirkind. Applied Optics. Vol. 4. 1965. p. 1077. and 
discussion with R. Zirkind. 11/88. 

AO 236 of 6/61 for BAMIRAC had been set up with the University of Michigan P^.^^ously for a br^ 
scxoTresponsibmcs connected with data for ballisUc missile defense largely m the infrared. 
Amendment # 2 was for $8.3M. 


construction and shakedown were expected to be completed, it would be mmed over to the 
University of Hawaii.** 

Soon after these initial steps by ARPA. a directive arrived from Harold Brown, 
then DDR&E, giving space object identification (SOI) and tracking a high priority in DoD. 
Since AMOS' capabiUties were designed for this purpose, its funding was increased. The 
University of Michigan undertook the design of two 48-in. infrared telescopes, on a 
common mount and shaft, one mainly for tracking and the other for special observations, 
and of a 60-in. telescope separately mounted, mainly for work in the optical spectrum. 
Design was completed in 1963 and construction of the foundation and buildings 
commenced by the Army Corps of Engineers.^ The Corps constructed the entire facility 
except for telescopes and domes. The three high quality mirrors were completed to 
diffraction limited tolerances, successfully and at quite low cost Special coatings were 
added to the IR mirrors to enhance reflectivity over the 1-30 micron range. Telescope 
mounts were of cast steel, a bit unusual, since most astronomical mounts involve welded 
pieces. This decision was made by ARPA, and the risk accepted to reduce costs. 
Successful casting saved SIM.^ The bearings were formed with very close tolerances, in 
order to allow the desired pointing and tracking accuracy of - 1" arc at angular rates 
required to track satellites and reentry objects. No telescopes of this size and weight had 
previously been constructed to the tracking specifications of AMOS.' However, the only 
hitch that developed in the construction occurred in the domes, which also had to have 
rapid motion capabiUties, something new for such structures. A separate contractor made 
the first domes, but tiiese were found to vibrate excessively. The previously helpful 
astronomers pitched in again to correct the problem.8 Considerable re-woric was involved, 
which caused an overrun, in mm forcing cancellation of plans for advanced 
instrumentation, which included, in 1964. an interferometric spectro-radiometer and 
computer-controlled articulated minors.^ 

4 The initial lease was for 25 years from the University of Hawaii, beginning in 1963, R. Zirkind. ibid. 

5 AO 389 of 8/62 and 482 of 5/63 to the Army Corps of Engineeis. 

6 Discussion with R. Ziridnd 11/88. 

7 nie Baker-Nimn satelUte tracking camera was ^nailer and Ughter with JO^Pfr?: "^^^^ 
tracking accuracy of about 2". "The Bakcr-Nunii SateUite Camera by K^^^^^ 
Telescope, Vol. XVI, Jan. 1957, p. 3. This system also had several successes m SOI, see e.g„ Hayes, 
loc. ciL.p. 121-2. 

8 A.MeineloftheUniversityofAri2onawasparticularlyhelpful. Discussions with R. Zirkmd U/88. 

9 R. Ziikind, ibid. 


Construction of AMOS was completed by 1967. Between then and about mid- 
1969 there was an initial phase of evaluation, caUbration and testing of the telescopes' 
computer control and tracking algorithms, and of the associated infrared arrays, 
radiometric, photometric and imaging equipment. A data link with a radar at another 
location in the Hawaiian area was estabUshed, to facUitate tracking.!** As originaUy 
envisioned, astronomical objects were used for caUbration. Initial attempts were made with 
some success to acquire and track satellites and otfier space systems. An early success was 
a photograph and tracking of one of NASA's APOLLO modules." 

Figure 1, from a current Air Force brochure.i^ shows picmres of the telescopes, 
housed in the largest dome shown in Fig. 2, which also exhibits other features of the 
AMOS facility as it is today. The optical systems provided for several instrument mounting 
platforms for different detection and imaging systems. Botii IR and optical systems had 
long focal lengths to allow fine image definition. 

A second data link witii a tracking radar on anotiier island was estabUshed, and this 
and other radars were reUed upon, togetiier with information from the NORAD network for 
initial tracking inputs. A low-power ruby laser was also installed, as a first step toward a 
laser radar target illumination technique. 

By 1969 die quaUty and potential of AMOS had been demonstrated and a second 
phase began in which die Air Force becanoc die ARPA agent Ue Air Force also began to 
support projects to measure properties of reentry bodies at the faciUty under its ABRES 
project The University of Michigan was replaced, as AMOS manager and operator, by 
industrial contractors, AVCO and Lockheed,i3 Computer and software advances further 
improved tracking capabiUties. In the early 1970*s advances in semiconductor state of die 
art allowed a much improved, larger infrared sensor array to be combined widi a contrast 

1 0 AMOS Advanced ElectroOptical Program, RADC 111-86.215. Feb. 1987, p. 2. This tcpon 
a brief history of AMOS since 1963. 

1 1 Discussion with Glen Rogers, AMOS, 1 1/88. 

12 AMOS/MOTIF brochure, undated. 

1 3 A.0. 2320 of 1 1/22 and RADC, ibid 



Figure 2. AMOS/MOTIF/GEODSS Observatory Buildings 
photometer and television camera in an "Advanced Multicolor Tracking" system. A higher 
power ruby laser was designed and instaUed to woric with one of the infiaied telescopes, to 
conduct initial ranging experiments. These improvements allowed IR and visible 
measurements to be obtained on reentering vehicles and penetration aids of the Minuteman 
Series and on several satelUtes.^^ Assistance was also provided to NASA to help with 
problems on the SKYLAB. 

14 RADCibid. 


In the late 1970's successful space object measurements continued in the infrared 
and visible, and laser ranging and Ulumination experiments began.^^ Eventually, a 
dedicated laser beam director was constructed. Preparations began for the installation of 
the ITEK compensated imaging system {CIS) which had also been developed by DARPA, 
to be used with the 60-in. telescope on low-altitude space objects because of the limited 
effective field of view.i* A number of measurements of high atmosphere turbulence 
related to CIS performance were made. Precision tracking improvements continued, 
particularly in characteristics affecting hand-off to local and distant tracking systems. 

A higher power CO2 laser was installed and used for experiments for ranging and 
illumination of more distant objects. In 1979 AMOS' twin infrared telescopes and 
associated systems became part of the Air Force Space Track Network and was renamed 
MOTIF: Maui Optical Tracking and Identification Facility. 

In the early 1980*s DARPA-supported AMOS activity included more detailed 
measurements of background, high cirrus cloud properties and aunospheric turbulence. 
Measurements were made on meteor trails in the infrared, and on the core of the M-87 
galaxy in the visible." Atmospheric compensation experiments began using Lincoln 
Laboratory dcformable mirror technique for directing a laser through the turbulent 
atmosphere. Several supporting experiments have been made for SDI in the atmospheric 
infrared windows.^^ The compensated imaging system was tested and installed on the 
60-in. telescope, A LWIR capabiUty was also added to the 60-in. on a side mount, and the 
60-in. muror was coated to improve its IR reflection. 

By 1984 AMOS had become a highly automated system, and DARPA transferred 
AMOS to the Air Force. RADC is now responsible for AMOS* R&D and the Air Force 
Space Command for the operation of MOTIF. A summary of current AMOS-MOTIF 
capabiUties is routinely issued by the Air Force. SDI now supports a substantial fraction of 
AMOS' activity. 19 

15 E.g.. AO. 2837 of 7/74. 

16 A description of this Itck system is given in the chapter on "Adaptive Optics," by JJi. Vyc 
Hardy. Chapter 8, p. 101 of Arms Control Verfication, Pergamon 1986. 

1 7 Direct Infrared Measurements of Thermal Radiation From the Nucleus of Comet Bennett, by 
Myer. Ap. L.. V. 175, 1972. p. U9. 

18 RADC, ibid. 

19 SummaryofAMOS-Technical Activities- 1987. RADC TR.87.301. May 1988. 


One of the original objectives for AMOS, astronomical infrared research, has been 
canied out only to a very minor extent 20 However, academic IR astronomy is now 
beginning to flourish with several telescopes in the U.S. and also at Mauna Kea (near the 
active volcano). What has caused this area to bloom is the availability of larger IR focal 
plane arrays, developed largely with DARPA support. Some of these arrays had been 
tested at AMOS.^^ 

Suggestions have been made by some members of the astronomical community, 
notably the Meinels (who have been involved with AMOS from the beginning) to begin 
planning for larger (10-meter range) aperture. computer-controUed. articulated mirror 
telescopes for the next-generation AMOS.^^ 


AMOS was an ARPA initiadve to construct an astronomical-quaUty facility for 
observations of sateUites and for astronomical research. The Air Force had used the IGY's 
Baker-Nunn telescope-camera for satellite observations, but AMOS was to be a larger, 
more complex and heavier telescope, with angular tracking quaHty at least as good as the 
Baker-Nunn. The step to construct AMOS was considered risky at the time, but not 
excessively so by competent astronomers, who were interested enough to provide help 
with design at the early and later stages of the project The sudden increase in priority for 
sateUite observation techniques enabled AMOS construction and use to proceed quickly. 
An academic contractor. University of Michigan, buUt the telescope. Initial plans were to 
turn AMOS over to the University of Hawaii, after ten years operation. After its 
construction, however, operational use of AMOS became predominant, and the plans for 
academic uses were on the one hand awkward, and on the other hand academic groups 
were, at the time, distancing themselves from miUtary-related programs. Industrial 
operation of these facilities was therefore considered more appropriate. 

Over a nearly 20-year period AMOS has met its primary objective of serving as a 
unique faciUty for electrooptic R&D and operational use. and is now considered a national 
asset. During this time many advances in electrooptic and related technology developed by 
DARPA have been efficiendy tested and used at AMOS. A key feature was tiiat 

20 Discussion with James Myers, Photon Research, Inc. 11/80. See Fn. 17. 

21 Sec e.g.. -Astronomical Imaging With Infrared Array Detectors,- by 1. GaUcy. et al.. Science. Vol. 
242, 2 Dec. 1988, p. 1264. 

22 "Summary of AMOS Technical Activities 1987" ibid., p. 16. 


astronomical objects of known brightness and spectral characteristics could be used for 
caUbration purposes. The success of AMOS is attested to by its past and current use for 
reentry and penetration aids studies by the Services and SDI. and as a part of the AF Space 
Track Systems. While DARPA suppon is now in the mode of support of "users," the 
challenges in the operational areas do not seem to have diminished 

While the original objective for AMOS also included astronomical research, this has 
occurred only to a very minor extent, for reasons outline above. AMOS, however, has 
been a unique test bed for focal plane arrays developed by DARPA, which have made a 
substantial contribution to the presently blooming field of IR astronomy. 

•After its initial demonstration of operational capability, transfer to the Air Force 
occurred gradually. The Air Force has collocated at the AMOS facility three of its 
GEODSS systems, developed also partly with DARPA support,23 to automatically detect 
and track satellites at geosynchronous distances. 

The initial AMOS facility cost appears, from project records, to have been 
approximately $12M. The cost of the later phases, including operations and improvements 
such as the CIS, and suppon of AMOS operations for some DARPA R&D projects, 
appears to be about $90M. 

23 AMOS user's manual, RADC. 


























The Air Force Maui Optical Station (AMOS), and the Maui Optical Tracking and Identification 
Facility (MOTIF) are co-located at an altitude of 10,000 feet on the crest of Mt. Haleakala, 
located on -the island of Maul, Hawaii. This high altitude location is characterized by a 
relatively stable climate of clean, dry air. The low levels of particulate matter and absence 
of significant scattered light from sea-level sources provide excellent conditions for the 
acquisition and viewing of space objects. The facility was constaicted during a two year 
period beginning in l963. During the past twenty years, the site has evolved to its present 
configuration, which includes four primary optical testbeds: the 1 .6-meter telescope, the 
dual 1 .2-meter telescopes, the Laser Beam Director (LBD), and the Beam Director/Tracker 
(B D/T) . These four optical telescope systems, along with the facility's sensors and computer 
resources, form the basis for both the Air Force Systems Command's (AFSC) AMOS 
Program, and for the Air Force Space Command's (AFSPACECOM) Spacetrack MOTIF 
program. Both organizations share the facility. AFSPACECOM maintains and operates 
the site as facility host; and AFSC. through it's executive agent, the Rome Air Development 
Center (RADC), is the tenant supporting measurement programs, special testing, and 
visiting experiments. 

The AMOS 1 .6-meter telescope is one of the finest optical instruments of its size in the 
world. In the absence of atmospheric-induced image distortion, the telescope permits 
diffraction limited performance (approximately 0.1 arcsecond resolution, or 1 ft. at a distance 
of 500 miles) at all mount attitudes above the horizon. The clear aperture is 1 .57m and the 
effective focal length is 25m. Broadband mirror coatings (Al plus an SiO overcoat) allow 
spectral coverage from the visible through the LWIR. The telescope is attached to an 
equatorial mount on an azimuth turntable. The mount has hydrostatic bearings, 23-bit shaft 
angle encoders on each axis, and is servo-driven by direct current torque motors under 
control of a Harris 500 computer. This system allows absolute pointing to ±2 arcseconds 
and tracking to ±1-3 arcseconds (depending on target velocity) at tracking velocities up to 
2 degrees/sec and accelerations to 2 degrees/sec^. An acquisition telescope with three 
switch-selectable fields of view is mounted piggyback on the north face of the 1 .6-meter 

Two instrument mounting surfaces are available for sensor packages on the 1.6-meter 
telescope. The rear surface is currently dedicated to the Compensated Imaging System 
(CIS), an adaptive optical device that compensates in real-time for atmospheric turbulence- 
induced distortion of satellite images. The side surface supports a sensor package which 


currently includes the Enhanced Longwave Spectrometer/lmager (ELSl), which is a dual 
infrared acquisition, imaging array, and the AMOS Spectral Radiometer (ASR). which Is a 
26 detector element MWIR/LWIR radiometer. An 8000 element Platinum Siliclde (PtSi) 
infrared Charge Coupled Device (CCD) is also included for infrared imaging in the 3-5 
micrometer spectral band. A sensitive Intensified Silicon intensifier Target (ISIT) Camera 
is also present in the package. 

The AFSPACECOM 1 .2-meter telescope complex represents a unique capability which 
functions as a fully integrated sensor In the Spacetrack Network. Two 1 .2-meter telescopes 
are mounted on opposite sides of a single polar axis, and are fixed to a common declination 
axis. The mount shares the same operating systems and performance parameters as the 
1.6-meter mount. Both 1.2-meter telescopes are classical Cassegrain optical systems, 
having parabolic primaries and hyperbolic secondaries. One telescope (B29) has a back 
focal distance of 29 inches, a relative aperture of f/20, and a focal length of 24.5m, while 
the other (B37) has a 37 inch back focal distance, a relative aperture of f/l6. and a focal 
length of 19.8m. Bctfi telescopes have primary mirror support systems which incorporate - 
air bags for axial support and mercury filled belts for radial support. An acquisition telescope 
is mounted piggyback on the 829 telescope. 

There are three mounting surfaces on these telescopes, one on the B29 telescope and two 
surfaces on the B37 telescope. The 829 houses the Advanced Multicolor Tracker for AMOS 
(AMTA), a square array of 25 cooled Cadmium-doped Germanium (Ge:Cd) detectors. The 
sensor is fitted with seven remotely programmable spectral filters that operate in the 3-22 
micrometer band. The system is used to collect kjw dispersion infrared spectral data pn 
targets of Interest, and to perform manual or closed-loop tracking of non-solar Illuminated , 
targets. Sharing the light beam with AMTA is the Contrast Mode Photometer (CMP), whlch- 
provides visible photometric signature data simultaneously with AMTA infrared signatures. 

The rear instrument surface of the 837 telescope houses the Low Light Level TV (LLLTV) 
Package, for detecting and imaging resolved targets, and for detecting very faint, unresolved 
deep space objects. The LLLTV consists of a high-gain, astronomical quality Intensified 
SIT camera with narrow and wide field of view optics. The package also contains a 1 6 mm 
cine camera for a classical imaging capability. The camera has a variable frame rate (2-1 00 
frames/sec), a tri-mode shutter providing consecutive exposures in the ratio of 1 :3:9, and a 
filter wheel for color spectral filters. The side instrument surface of the B37 houses an 
atmospheric turbulence measuring device, and additional mounting space is available for 
visiting experimenters. Mounted on the 837 telescope housing is a small 1 Joule pulsed 
ruby laser used as a Cirrus LIDAR Probe (CUP), and an 18 inch receiver telescope is used 
to detect backscattered light from the atmosphere. 

The Laser Beam Director is an optical system which provides precise laser beam pointing 
and tracking. The system utilizes a series of foced mirrors and beam expanders to take the 
output of a laser system, expand it to 24 inches, and direct it to a 36 inch azimuth/elevation 
gimbaled tracking mirror, from which It is projected Into the atmosphere. The 24 inch beam 
expander and the 36 inch tracking mirror are mounted on an azimuth turntable which is 
locked prior to a tracking operation. The LBD has supported the AMOS pulsed ruby laser 
system, a three stage Q-switched and conventional mode laser producing pulse energies 


of about 8 and 80 Joules, respectively, for laser ranging and illumination of objects in space. 
The beam director has been designed to enable user agencies to mount the.r own laser tn 
the sub-dome area and utilize the existing optics and pointing to conduct measurement 
programs tailored to a specific laser system. 

The newO;8-meter- Coude Beam Director/Tracker is a versatile system that can acceptup 
to a 15 cm. beam from a variety of lasers, and project it to an object being tracked. The 
beam may be projected from the BOrr without expansion, or be expanded up to 0.6 meters. 
In addition to the Coude path, the system includes a Cassegrain mounting surface. The 
BDT mount is an altitude-altitude configuration with a Coudfe path to bnng the laser beam 
to the projection optics from a fixed point on the observatory floor below. The mount can 
track at velocities up to 5 degrees/sec and angular accelerations up to 4 degrees/sec . The 
son is operated with a variety of lasers, including systems installed by visiting user 
agencies. The LIDAR Acquisition/Sizing Experiment (L^SE) system is currently in use with 
the BD/T This bistatic C02 laser transceiver is designed to provide measurements of target 
range and range rate at ranges in excess of 2 Megameters. independent of time of day. 
The system was designed to serve as an experimental test bed for precision dynamic 
measurements. Doppler imaging and micro Doppler measurements. 

In addition to the large optical systems and sensor capabilities at the AMOS/MOTIF site, 
extensive computer facilities have been installed as well. The Mount Control System (MCS) 
Harris 500 computers direct the operation of the 1.2-meter, 1.6-meter. LBD. and BDfT 
mounts. The MCS allows each mount to independently acquire and track targets with a 
high degree of precision, and to employ data from remote sensors, such as off-site radars, 
to achieve acquisition when necessary. In addition, two MODCOMP computers provide the 
capability for collecting, recording, displaying, editing, processing, and transmitting 
AMOS/MOTIF data. One MODCOMP is part of the Data Transmission System (DT5). 
which is capable of simultaneous, real-time acquisition and storage of metric. Photometric, 
and infrared data. The second MODCOMP is part of the Communication System (CMS), 
which takes information from the DTS and formats and transmits the data via AUTODIN to 
AFSPACECOM. Other computers at the facility perform digital image storage and trans- 
mission, data analysis, and database management at the site. 

Extensive support systems exist at the site to operate and maintain the complex and unique 
optical systems and sensors at AMOS/MOTIF. These include a satellite-based Global 
Positioning System (GPS)-referenced timing system, secure 2400 BAUD worldwide 
AUTODIN, and a secure voice system. A separate support building adjacent to the obser- 
vatory facility contains a mirror re-coating laboratory with a vacuum tank capable of holding 
the telescope primary optics. The support building also houses a machine shop, electronics 
shops, welding shop, carpentry shop, and parts storage. 





The VELA HOTEL SateUitcs were part of the ARPA VELA program assigned by 
DoD.i The objective of the VELA HOTEL piojea was to develop sateUite technology and 
global background data to detect nuclear explosions taking place in space, and eventually 
also in the eanh's atmosphere. The first such experimental satellites were launched in 1963 
and were very successful, with pcrfonnance, cost and lifetime far better than expected, 
which allowed progressive improvements to be made rapidly in the detection systems and 
related sateUiie technology. This success also provided interim monitoring capabUity in 
support of the Limited Test Ban Treaty in 1963. banning nuclear tests in the earth's 
atmosphere and in space. In 1970, after six VELA HOTEL SateUite pairs had been 
launched without failure and operated successfuUy in orbit, the program was taken over by 
the Air Force. The cunent Air Force operational nuclear test detection system includes 
improved detectors of the type developed in the VELA Hotel program, incorporated into the 
GPS/NDS integrated navigation and nuclear e3q)losion detection satellites. Six of a planned 
constellation of 18 are. so far, in orbit Tlic VELA-typc instrumentation in die HOTEL and 
later satellites have been credited witii detecting: "every nuclear event set off above ground 
that it has been in a position to see."^ 


In May 1959, the High Altimde Detection panel (Panofsky Panel) of the President's 
Science Advisory Committee, recommended a sateUite system be used to detect nuclear 
tests in space and in the atmosphere, as part of the overaU basis for verification of a funire 
nuclear test ban treaty. This panel also considered it possible, but difficult, to hide even 

VELA means watchman in Spanish. Hotel was apparenUy, not an acronym. Other partt ^^^^ 

program were: VELA UNIFORM, detection for underground explosions, and VELA SIERRA tor 

ground-based methods to detect nuclear explosions in the amiosphere and m space. 

"Satellite VerificaUon of Arms Control Agreemenis." Harold V. Argo in Arms Control Verificathn. 

Pergamon Press, 1985, p. 292. However, an apparenUy <^°"»'°^?^^li"f 

in Sept 1979. See "Monitoring The Tests." IEEE Spectrum, July 1986. p. 63-64. and Alvarez, oy 

L.W. Alvarez, Basic Books, N.Y, 1987, p. 249. 


smaU nuclear tests in space. To succeed in this would require special measures, such as 
hiding detonations behind the moon, using heavy lead shielding, or conducting the tests at 
very great distances. Technical Working Group I of the Geneva Conference on 
Discontinuance of Nuclear Weapons test recommended, in July 1959, "placing five or six 
large satclUtes in earth orbit at a distance of 180,000 miles to detect radiations from nuclear^ 
explosions in space." The satelUtes would be supplemented by special equipment placed in ^ 
the 170-odd ground-control points of the recommended Geneva system for monitoring 
nuclear explosions underground and in the atmosphere.^ 

ARPA was assigned overall rcsponsibUity by the President, in late 1959, for project 
VELA, aimed at developing technology for detection of nuclear tests and veiification of a . 
nuclear test ban treaty. ARPA began immediately to plan for the required launchers for ; 
VELA HOTEL, the space segment of VELA, and with the assistance ofnh^ AEC • 
laboratories at Los Alamos and Sandia, design of a satelUte system commenced in^.the 
summer of 1959.* 

As prescribed by the Geneva Technical Working Group, earth-based technologies. -: 
to detect nuclear explosions in space were also investigated under the VELA SIERRA. ' 
ground-based nuclear detecuon project, including an optical system to detect air' 
fluorescence caused by X-rays,^ nuclear-burst-caused ionospheric effects on VLF radio ^ 
propagation and absorption of cosmic radio noise.^ 

Some felt that the costs of an adequate satellite system could be very hjg^^ 
particularly if the possibility of lead shielding of X-rays from the explosion and ot%^^^-' 
possible evasion methods were taken into account, along with the lack of relevant 

3 Kennedy, Khruschev and the Test Ban, by Glenn T. Scaborg. U. CaL press 1981, p. 19. 

4 AO 102 "VELA" of 9/59 to Sec. AF for nearly $70M. and AO 140 "Project VELA" of 4/60 to AEC, 
$4.4M.' The AEC labs had already been working on the problem with AEC suppon. iw. 
Developments in the Field of Detection and Identification of Nuclear Explosions, Summary of Hcarmg 
on July 25-27. 1961, Journal Committee on Atomic Energy, Apnl 1962, p. 5. 

5 Ground-based opucal systems for detecuon of nuclear explosions in f ^ce were apparenUy^fiel^^^^ 
and used beginning in 1961. but were, initially. ^^J^^'J^^y- 

Fluorescence Detection System." by DJi. WcsierveU and H. HoerUn. Proc. IEEE, VA 53. #12, 1965. 
p. 2078. . u 

6 OTH radars to detect nuclear explosions in the ionosphere were proposed by the U.S. but r^je^d by 
Se Soviet Union. See testimony by W. Panofsky. in "Technical Aspec^^ o De^f " tSSi CoS ^ 
of a Nuclear Weapons Test Ban." hearings before a Subcommittee on Radiauon JCAE. 86ih Congress, 
2nd Session, April 1968, p. 48. *^ ' 


background data and the possibility of unreliabUity of the space systems invoivedJ 
Because of the controversy, a joint agency technical group was set up by ARPA to plan and 
steer the VELA HOTEL project, with AF Space Division chairmanship. 

A number of instruments were also flown piggy-back on other early U.S. Defense 
and NASA satellites to test instrument performance and make preliminary background 
measurements.* Estimates were soon made that 3 to 5 launches of sateUites, in a five-year 
program, would prove adequate for defining a prototype system.^ Detection experiments 
were also performed by launching rockets from Hawaii during the 1962 high-altimde 
nuclear test series. 1° Under die DARPA program six pairs of VELA sateUites were put into 
orbit, the first pair in 1963. and the last pair in 1970. Table 1 gives a summary of the 
launch dates, and information on the satellites' equipment and stabilization. 

Table 1. VELA HOTEL Satellite Launches 

Satellite Pair 

Dale In Orbit 

Detection Equipment 



16 Oct 1963 

Nuclear (space explosion) 

Spin (fixed axis) 


27 July 1964 

Nuclear (space exptosion) 

Spin (fixed axis) 


20 July 1965 

Nuclear. Bhangmeter 
(atmospheric explosion) 

Spin (fixed axis) 


28 April 1967 

Nuclear. Bhangmeter 

Earth-oriented (gravity) 


23 May 1969 

Nuclear, Bhangmeter 

Earth-oriented (gravity) 


8 April 1970 

Nuclear, Bhangmeter 

Earth-oriented (gravity) 

7 See A Scientist at the White House, by G. Kistiakowsky. Harvard, 1976. pg. 76 and "Scientists and 
Politicians," by R Jacobson and E. Stein, U. Mich. Press, 1960, pp. 191-2 

« Some early results arc described in Uic tcsumony of Dr. A. Schardu ARPA Vela Hotel program 
manager and "Developments in Technical CapabiUiy for Detecung and Ideniifymg Nuclear Weapons 
Tests," hearing l)efore the JCAE, 88th Congress, Isi Session. 1963, p. 331. 

9 Schardt, ibid., p. 321. 

1 0 Seventeen rocket pay load measurements were successful out of seventeen launclied. See testimony of 
J^M H. Coon, in "Developments in Technical CapabiUues for Detecting and Idenufymg Nuclear 
Weapons Tests," hearings before the JCAE, 88th Congress, 1963, p. 390. 


The first pair of VELA satcUites were successfuUy launched in Oct 1963, spaced 
180 deg apart in a circular orbit at about 1 15,000 km^^ beyond die outer Van AUen Belt 
The second and third pairs were launched in July 1964 and 1965. All of these contained 
X-ray, neutron and gamim-ray detectors designed by the AEC Ubs,. which could measure 
the very characteristic signals of these types from a nuclear explosion. Figure 1 shows a 
photograph of the first two VELA HOTEL satelUtes mounted in tandem, and ready to be 
mounted on their booster rocket Each satelUte had an internal injection motor used to 
positionitinfinalcircularorbitof 115,000-kmradius.approximately ISOdegapan. These 
satelUtes had an icosahedral configuration, with cubic shaped X-ray detectors at each apex. 
The ganmia-ray and neutron detectors were inside. The second and later satelUtes carried 
instruments to measure background radiation to which die nuclear explosion detectors 
might be most sensitive.i2 Witii tiiis background infonnation, coincidences of multiple 
detectors of die same type and time histories of the different signal types could be used in 
the design of logic systems in the sateUite^^ to identify explosions with greater confidence. 
While the first VELA HOTEL sateUite detection payloads were constructed by Los Alamos 
and Sandia, the satelUtc frame, solar cells, etc.. had been built by TRW under a success- 
oriented performance incentive fee contract, one of die first of a long series of tfus type in 
the miUtary satellite business." Because of the exccllcni TRW pcrfoimance, a sizeable fee 
had to be paid by ARPA, which was done widiout objection, The Ufetime of rfiese first 
satellites had been expected to be nine montiis at most, but named out to be years. Taking 

1 1 The Limited Test Ban Treaty, including provisions against nuclear tests in space and in the atmosphere, 
had been signed before this, in April 1963. 

12 "The Vela SateUite Program for Detection of High Altinide Nucle^ Detonations " f -^S^ns^^* 
Proc IEEE. Vol. 53, 1965. p. 1935. "Vela SateUites Measurements of Particles in the So^ Wmd and 
the Distant Geomagnetosphere." by James R Coon, in Radiation Trapped in the Earths Magnetic 
Field, B, M McCormack, ed, Reidd 1966, p. 231-236. 

13 Considerable effort went into the design of the logical systems at Sandia because ^^^f^^.^^^ 
avoid false alanns. See Jacobson and Stein, ibid., p. 191. For the situauon as of 1965 s«^. A 
modular System of Logic for the Vela SateUite Program," by W. McGoldnck, et al., Proc, IEEE, Vol. 
58, 1965, p. 1959. 

Discussion with Dr. C. Cook, 12/89. 
15 Discussion with Dr. R. SpK)ull, who had been ARPA director at the time. l^^J- Def 
McNamara cited the VELA Hotel contract in his 1964 report to the President on Cost Reduction. TT^e 
success of this CPIF contract can be credited partly to the clear technical descnpnon of reqmremenis by 
ARPA. see Richard J. Barber Associates. DARPA History, ibid The success m later con^f this 
type ci be credited, in part, to their heavy "incentivation" possible due to the "special handling of the 
satellite program. Dr. C. Cook, ibid. 



advantage of the remarkably successful launch and successful payload performance 
together with lower costs and longer lifetimes, ARPA changed the schedule and payload, 
as things went on, to progressively incorporate improved nuclear detection systems. 

The test ban treaty of 1963 gave incentives to extend the satellites* capabiUty to 
atmospheric explosions. The multistation Geneva ground-based system was becoming 
appreciated as being very costiy and a large, difficult burden on the U.N. (or some other 
international body), and the satclUtcs offered a way to provide a substitute for the 
atmospheric detection role of these stations.*^ 

The key technology for this purpose was the "bhangmeter." a version of an optical 
instrument that had been used previously by the Los Alamos Laboratory for measurement 
of the Ught emitted by atmospheric explosions and proposed by the laboratory for this 
appUcation. In order for the bhangmeter to detect the characteristic optical signamre of 
nuclear explosions in the atmosphere, it was necessary to first use it to obtain some 
preliminary data on the brighmess background characteristics of the earth. The third 
sateUite pair contained a bhangmeter, but Umited earth background data was acquired 
because of the spin-stabilization then used. To detect nuclear explosions in space, no 
particular directional charaaeristics were required for the other instruments. 

When the fourth VELA sateUite was launched in 1967, space technology had 
advanced enough to allow its axis to be oriented towards the earth's center so that a 
bhangmeter looking downward could detect and measure the double-humped optical 
signamre characteristic of an atmospheric nuclear explosion, which could also be used to 
estimate yield.i^ The last two sateUites pairs of the VELA series also contained 
electromagnetic pulse detectors for nuclear explosions in the atmosphere. 

The early gamma-ray detectors, which like the X-ray detectors employed 
scintillators, were improved to have better time and spectral resolution and in 1967 the 
fourth pair of VELA sateUites detected, for tiie first time, gamma-ray bursts identified as 

16 Seaborj? ibid. p. 147. discusses the probable impraciicality of the Geneva Systems as first proposed. 
C^^!i^mt;^s were given by CNiBeyer of ARPA, in lesumony before J^-";^ ^^^^ 
Atomic Energy in 1963. See "Technical Aspects of Detection and ^"/'f ^ona^^^^^^ 
Weapons Tests Ban." Hearings before a subcommittee on Radiauon of the JCAE, 86th Congress, 2na 

Session, April 1960, p. 367 ff. 
17 Argo. ibid., (Ref. 1). p. 298. 


coming ftom distant coUapsing star events.i8 The sixth and last VELA HOTEL Program 
satelUte pair was put into orbit in 1970. Several of these satelUtes axe still operating. 

When the Air Force (at first SAMSO, and later AFTAC) took over the satellite 
nuclear detection responsibUity after 1970. the nuclear explosion detection payloads 
(beyond the existing VELA HOTEL systems) were at first combined witii otiier 
instruments, for reasons of economy, in geosynchronous sateUiies.!' since July 1983. die 
nuclear test detection responsibility has been given mainly to the GPS/NDS combined 
navigation and nuclear test detection satelUtes systems, planned for 18 sateUites at 20,200 
km altitude (within the outer Van AUen Belt) and 55 deg orbits, and now being built up as 
launch capabUities allow. Six arc presently in orbit. Most of the GPS/NDS systems 
include X-ray, bhangmeter and an EMP detector, as die VELA satellites did, and some also 
contain a dosimeter to assess damage to on-board systems and to detect magneticaUy 
trapped electrons and ions from a nuclear explosion. The recent GPS/NDS systems do not 
include gamma or neutron detectors, but this capability is apparendy still available on other 
satelUtes.20 xhe accurate timing inherent in die GPS system is used also for locating die 
source of signals detected by die X-ray. bhangmeter or EMP detector, aUowing correlation 
of die times of arrival at different GPS/NDS satelUtes. Signals received at a number of 
satelUtes are analyzed at ground stations for positive detection, identification, location and 
yield estimation of a nuclear explosion, useful not only for monitoring nuclear tests but also 
for wartime assesstnent of nuclear attacks. 


The VELA assignment was given to ARPA by die White House and DoD. A rough 
prescription of the technology involved was available from die Geneva Technical Working 
Group I and Uie Panofsky Panel. However, dicre was stiU considerable confusion over 
how much detection capability would be required, and at what cost Confidence was also 
not high, until about 1963, in launch success or in payload lifetime. In retrospect the 
VELA HOTEL satellites benefited very grcatiy from a combination of what was, at die 
time, an unusually successful launch series, togedier widi die high quality nuclear test 

18 -Gamma Ray Astronomy/ G. Ramairy and H, Ungenfeltcr, in Annual Reviews of Nuclear a 
Particle Science, Vol. 32, 1982, p. 242. 

19 Panofsky, ibid., where it is pointed out that beyond detection of a nuclear explosion, ideniificadon 
the test violator would need additional information from other surveillance sources. 

20 ARPA, ibid., p. 302. 


instrumentation and rigorous logic control technology avaUable ftom the AEC laboratones. 
The logical subsystem was considered very important, in order to give high confidence m 
any detection made by the sateUite. There were many technical risks: launchers and 
payload design, overall payload reUabUity and lifetime, and importantly, the radiation 
backgrounds on which information had to be built up over time. On the basis of the 
sequential accumulation of information on nuclear system performance and background and 
the rapid advance of space technology. ARPA's working group changed die technical 
specifications as the series went on. 

The main feanires of the nuclear components of the sateUite system to detect high 
altitude nuclear explosions were clear after three successM launches, as had been estimated 
after some background data had been attained. But the 1963 treaty banning nuclear 
explosions in space and in the amiosphere, and the high cost for die Geneva ground-based, 
multi-station system then under discussion for monitoring, gave strong incentive to have 
satelUte systems to detect atmospheric tests worldwide. TOs required new technology on 
the satellite, which again was available from previous AEC programs. Hie bhangmeter, an 
optical instrument developed previously by Los Alamos, was added to the payload. and 
sateUite technology now aUowed an earth orientation to look downward witii it. Addiuon 
of the bhangmeter for the detection of amiospheric tests required a new and different kind 
of background and discrimination logic. Proving out this technology required three more 
experimental payloads which again were successful. The phenomenal run of successful 
launchers can largely be credited for the success-oriented progress of die project 

The Air Force apparendy was impatient at first to take over responsibility, but 
evenmally recognized the cost savings in the project and in its CPIF contract with TRW.^i 
Some known risks to avoid, which would have required a larger number of detection 
sateUites and consequendy high costs, were accepted for economical and political reasons. 
The early sateUites' remarkable success provided an interim operational capabUity for test 
detection, and also for diagnostics and rough location of nuclear explosions occumng m 
die amiosphere. TTie experimental VELA HOTEL sateffite system was acmaUy operanonal 
for many years. When die Air Force took over, detection packages similar to diose m die 
VELA sateUites were combined, pardy for economy, widi oUier payloads on die Air Force 
geosynchronous satellites. These were in a different radiation environment from the 
VELA satelUtes. but information was available on Uiis background from "piggyback 

21 W Richard J. Barber ARPA History quoBS a letter from Gen. Shriever to this effect Ibid. p. V.32. 


experiments on other geosynchronous satellites. Apparently the ARPA program 
envisioned eventual use of its pixxiuct in the geosynchronous satellite 22 Now a somewhat 
modified version of the VELA HOTEL system is earned on the GPS/NDS satelUtes, which 
provides a wartime attack and damage assessment capabiUty as well as nuclear test 
detection and location. 

ARPA expenditures for VELA HOTEL, from available records were approximately 
$150 miUion, including six launches, payloads, and data analysis. The incentive contract 
to STL was estimated to have saved $26 milUon.23 Expenditures for the successive 
generations of detection systems, including ground stations, from the early 1970s through 
the GPS/NDS, arc estimated as about $2 billion.^* 

22 Discussion with Gen. H. Dickinson, 7/88. 

23 Ibid. p. 29. Sec icsumony of Dr. A.W. Schardu in "Developments in Technic^ ^pa^^^^ 
Dete^rtion and Identifying Nuclear Weapons Test," hearings before the JCAE, 88th Congress. 1st 
Session 1963, p. 322. 

24 Dr. C. Cook. ibid. 






















1959 I 

s^n'c^a— T 

- - -ayload" "V^^i^^S^ ■ ■ 






GROUP X ^ m ^ mm 





VHI ■ 
^0/63 I 





















* 1968-77 








1 -29-90-1 M 




As part of project VELA, assigned to ARPA by the Secretary of Defense in 1959, 
VELA UNIFORM was a program of research in seismology and other techniques toward 
improvements in the detection and identification of underground nuclear explosions. As 
one of its first activities, VELA UNIFORM set up the first worldwide network of standard 
seismograph stations, the WWNSS, which has had a very great impact on seismology and 
its applications to our understanding of earthquakes and to geology, as well as to the 
problem of detection and identification of underground nuclear explosions . 


In 1958 an international committee of experts met in Geneva to define technical 
characteristics of a control system to monitor a possible nuclear test ban.i However, 
seismic data from ongoing underground nuclear tests in the U.S. soon indicated that the 
capabiUties of the system recommended by the Geneva Experts was considerably less than 
they had estimated. In the same period the "decoupling" theory was put forward, 
according to which a large explosion in an underground cavity could appear to be much 
smaller to a distant seismic monitor. These events led, in early 1959, to the formation in 
the U.S. of the Berkner panel on seismic improvement, which was asked to review the 
situation and recommend what changes would be needed in the Geneva system to bring its 
capabilities more nearly to the level the experts had originally estimated The Berkner panel 
recommended several such improvements in March 1959, and in a special report 
emphasized the urgent need for, and ouUined the desirable content of, an accelerated 
research program in seismology to better deal with the problems of detecting and 
identifying underground nuclear explosions.^ 

1 -VELA Ovcrview-ihe Early Years of the Seismic Research Program." by CJF. Romney. in "The 
VELA Program," DARPA 1985. Vela in Spanish means "watchman." 

2 The Need for Fundamental Research in Seismology," report of the Panel on Seismic Improvement, 
U.S. Department of State, 19S9. 


The recommendations and the rather comprehensive outline of needed research in 
the report of the Berkncr panel led to and guided the early stages of ARPA's VELA 
UNIFORM program, estabUshed in Sept 1959.3 One of the first steps suggested by the 
Berkner panel's repon was to equip, as soon as possible, selected seismographic stations 
worldwide with a standard set of seismographs, and equipment for accurate time and data 
recording, together with a central data repository.* ARPA, which was not strong in the 
sdsmology area at the time, depended on AFTAC, the Air Force Technical AppUcations 
Center, which had been active in the nuclear detection and seismology area since 1946. and 
had developed a detailed plan along the lines of the Berkner panel repon.5 aRPA then 
proceeded to implement this plan, one important aspect of which was assigning the task of 
installing the equipment and managing the WWNSS and its central data repository to the 
U.S. Coast and Geodetic Survey (USC&GS), an agency which had been involved in 
seismological activity for some time and was known worldwide.^ ) 

The USC&GS undertook the task with enthusiasm.The WWNSS instruments were 
to become the property of the stations or institutions in the different nations where tiiey 
were installed, and voluntary cooperation in data exchange, as had been die custom in 
seismology, was assumed. A committee of the National Academy of Sciences assisted the 
USC&GS on the choice of instruments and the selection of recipients.'' Proven, reliable 
instruments were recommended, one short and one long period type, each measuring three 
components of motion. Direct Ught-beam photographic recording was used. A single 
contractor, the GeoTechnical Corporation, suppUed the instruments for the 120 stations 
distributed around Uie worid. This was the first relatively large-scale industrial 
seismological instrument production of its kind. Figure 1. from Fairell.* shows a picture 

3 A.O. 104 of 9/59: "Vela Uniform." to AFTAC. 

^ Frank Press and David T. Griggs. "Improved Equipment for Existing Seismic Stations." Appendix I of 
the Beriaier report, ibid., p. 17 and 18. Besides making a very great improvement m seismology. « 
was envisioned that the distribution of seismographs could make it possible for other nauons to 
identify attempts at cheating on the test ban. Discussion with R. Sproull, 10/89. 

5 IDA TE 212 of Dec. 2. 1959: "AFTAC Development and Funding Plan: VELA." by R.S. Warner and 
F.C. Hazcn. 

6 A.O. 173 of 9/60 to USC&GS. 

7 "Specifications for a World-Wide Network of Standardized Seismographs." a report by the Committee 
on Seismological Stations, National Academy of Sciences, Washington, D.C., June 1960. 

8 W£. FaneU, "Sensors. Systems and Arrays: Seismic Instrumentation Under Vela-Uniform," in The 
Vela Program. ARPA 1985. p. 489. 


of one of the WWNSS systems, and Figure 2, from OHver and Muiphy,^ indicates the 
station locations. Each station was suppUed with a standard crystal controlled clock, and a 
radio system to receive and record time signals. Provision was also made for periodic 
calibrations of the WWNSS systems. A Seismology Data Center to copy and distribute the 
data was formed first in Washington and later in Asheville, N.C, under the USC&GS and 
finally in Boulder, CO under the U.S. Geological Survey. 

The WWNSS* installation involved many problems, technical, logistical and 
politicaLio The installation was essentially con^lete by 1963, with over 100 stations in 54 
countries, at a cost of about $9 milUon. The only notable non-recipients were Canada, 
which agreed to share data from their own system, and the Soviet Union. 

The WWNSS transformed seismology and became the main source of data for that 
science. In a 1979 National Academy Report, seismologist Jonathan Berger describes the 
impact of WWNSS (orWWSSN):" 

Until the mid-1960's a seismologist had to rely on a diverse set of 
seismograms that he had culled from various organizations and mdividuals 
throughout the world. Network analyses were, at best, extremely tedious 
and usually impossible, because in many, even the most rudimentary 
calibration (which way is up?) was unknown. With the deployment 15 
years ago of some 120 stations of the World Wide Standardized 
Seismograph Network (WWSSN), a large quantity of graphically recorded 
seismic data became available to die world's seismologists. 

When the WWSSN was established in the mid-1960*s, die world's 
intermediate and larger earthquakes were routinely and accurately located, 
and it was soon discovered that the vast majority of eanhquakes were 
confined to narrow zones spanning the globe. Further, certain parameters 
describing die source could be established. Using the model of an 
earUiquake as a fracture of the rocks over a plane, scientists could determine 
die orientation and direction of motion on this plane. This seismological 
evidence, on a global scale, contributed significandy to die development of 
the theory of plate tectonics in the late 1960's. 

9 J. Oliver and L. Murphy. "WWNSS: Seismology's Global Network of Observing Stations." Science 
V, 174, 1971. p. 257. 
Oliver and Murphy» ibid. 

1 1 "Impact of Technology on Geophysics," Nauonal Academy of Sciences, Washington, D.C. 1979. 
p. 65-66. 


Figure 1. The WWSSN 

(from Farrell, Fn. 8) 

Figure 2. Worldwide Network of Standardized Seismograph Stations 
Established as Part of the Nuclear Test Detection Program (USC&GS) 



According to OUvcr and Muiphy. the WWNSS airived "just in time" for the new 
development in geological concepts: 

In pan, the success of the WWNSS has resulted from the increase in the 
quantity, quality, and means for distribution of the data. To some extent 
successes occurred because the new data became available at the "right" time 
in history, just when die concepts of sea-floor spreading, continental drift, 
and plate tectonics were appearing, or reappearing, and undergomg 

The very earliest stages of the development of the sea-floor spreading 
hypothesis depended in only a limited and secondary way on seismology, 
for it was geomagnetism that held the key. Seismic activity was used to 
map the spreading zones, but the linear magnetic anomalies were die source 
of infomiation on spreading and rates of spreading. Very shortly, however, 
the contributions of seismology grew in imponance, and this discipline was 
able to play an important role in the testing and development of the 

Providing from three to five times as much data as previously available, data 
of much greater reliability from standardized, calibrated instruments, 
WWNSS allowed a drastic clarification and improvement of the delineation 
of seismic activity, earthquake focal mechanisms, and seismic wave 
propagation. ^3 

A 1977 report of the National Academy statesi^"^ 

In a litUe more than a decade, the WWSSN significantly increased our 
knowledge of earthquakes and of the earth strucnire and dynamics, while 
performing its initial mission of providing basic scientific information for 
the detection and identification of underground nuclear explosions anywhere 
in the worid. These major scientific advances provide important new input 
toward solutions of such national problems as the monitoring of nuclear 
tests, earthquake hazard reduction, understanding the origin and location of 
minerals and geothermal energy sources and die siting of dams and nuclear 
power plants. 

Regarding the nuclear test monitoring problem, Farrell says, more specifically: 

The WWSSN project has undoubtedly delivered more seismograms to 
seismologists tiian all other networks combined...Although set up as a 
research tool for studying fundamental problenis in seismology, it c^ be 
argued that studies conducted on data from this single network have been 

1 ^ Oliver and Muiphy, ibid., p. 2S7. 

13 Ibid., p. 18. OUver and Murphy Uliistraie ihis progress with several examples, Ref. 9. p. 25 

14 "Global Earthquake Monitoring." National Academy of Sciences, 1977, p. "i- Chapter 
report outlines the history of seismological networks and ihe accomplishments of WWNSS- 

15 Farrell. Ref. 8, p. 487. 


comparable in imponance to that provided by all other seismic systems for 
the problems of source identification and yield estimation. 

DARPA continued to upgrade the technology of the WWNSS, notably toward 
being more "digital;* to complement its capabiUtics with other stations having different and 
improved instruments, and to arrange for central processing of the digital seismic data. 
Most of this was done through the U.S. Geological Survey (USGS). In the late 1960's, 
DARPA also sponsored the development and installation of 10 high-gain, long-period 
(HGLP) seismographs which were later augmented with shon-period instruments and 
outfmed with improved digital recorders, and managed by the USGS as a complementary 
pan of the WWNSS. In 1973 DARPA and the USGS jointly developed and deployed 13 
Seismic Research Observatories (SRO), which included a new broadband borehole 
seismometer and an advanced digital recording system.!^ 

Berger describes the important characteristics of this upgrade from the standpoint of 
nuclear test discriminadon.^'^ 

When established in the mid-1960's, the WWNSS was confined by the 
sensor and associated electronics principally to periods shorter than 20 sec. 
Later in the decade, Pomeroy and others at Lamont-Doherty Observatory 
developed the high-gain long-period (HGLP) instrumentation that 
successfully modified seismometers to extend their useful range to 60-100 
sec. An outcome of their swdies and those of odiers was the discovery of 
an optimum period at which to discriminate between nuclear explosions and 
natural earthquakes. Based on this knowledge, two global arrays of seismic 
instruments "tuned" to this period were deployed - the Seismic Research 
Observatories (SRO) network and the HGLP systems. 

In parallel with the upgrade of instruments in the field, and the increase of digital 
data in quantity and quality, a new seismic data center has been set up to process and 
manage this data for the benefit of both geophysical research and international data 
exchange for treaty support.^^ 

Since the beginning of the WWNSS. it has been recognized that^^ 

16 ARPA Order #2880 of 6^4. Cf. also "Seismic Research Observatories. Upgrading the Worldwide 
Seismic Data Network," by J. Peterson and N. Oisini. EAS. American Geophysical Union. 1977. 
p. 548. 

17 Berger, Ref. 9, p. 67. 

1 8 "Tools for Seismic Data Analysis and Management for Research and International Data Exchange," by 
Ann U. Kerr, in The Vela Program, DARPA. 1985- 

1 9 Seismographic Networks, Problems and Prospects for the 80\ National Academy Press, 1983, p. 7. 


...DARPA has been responsible for virtually all advances in global 
seismo graphic networks... 

However, the support required for the continued operation of the WWNSS has 
been precarious since about 1967 when ARPA funding for it ceased due to Congress ruling 
that earthquake research was irrelevant to the ARPA mission.^o The responsibility for 
WWNSS was then eventually transferred to the U.S. Geological Survey. A similar event 
for the GSDN. the global scismological digital network, occurred in FY 1979, and as a 
result these networks have been reduced in size somewhat. However, much scismological 
research supported by DARPA depends on data from the routine operation of the GDSN 
and WWSSN.21 

At the present time it seems likely that the National Science Foundation and the 
uses will have a dominant role in any future upgrading and operation of the WWNSS, 
and the construction and operation of a "next generation" digital network, linked via 
satellite. Such an advanced system wUl also consist, largely, of technology generated 
through DARPA support . 


ARPA was given the VELA program responsibUity by the White House and DoD. 
AFTAC, at the time technically much stronger in the underground and atmospheric nuclear 
test detection areas, had prepared a comprehensive plan to cany out the Berkner Committee 
recommendations. However, AFTAC was not given VELA responsibility, probably 
because of its more direct military and intelligence connections. ARPA used tiie AFTAC 
plan to help guide its initial activity. 

F. Press of the Berkner panel had put forward the idea that a global "standard" set 
of seismographs and recording instruments was needed for VELA, actually could be 
carried out inexpensively, and would be very beneficial to seismology. It was also 
envisaged that a worldwide distribution of seismographs could help other nations to 
identify attempts at cheating on the test ban.22 The Berkner panel recommended that VELA 
carry out this WWNSS project, and this was included in the AFTAC plan. A National 
Academy Panel was formed to provide technical specifics for guidance of the WWNSS 
project WWNSS depended entirely on international data exchange and cooperation of the 

20 Communications from Dr. E. Rechtin, 10/89. 

21 Seismographic Networks, ibid., p. 11. 


kind that had been prevalent in scisniic research. The U.S. Coast and Geodetic Survey 
(USC&GS) was an appropriate choice of agent in view of its international connections. 
The C&GS had both recognized expertise and enthusiasm, and did a remarkable and 
difficult job in installing WWNSS and shepherding it through its early stages. 

WWNSS involved proven technology. The risk was in whether the network, 
based on an expansion of existing seismological voluntary practices, would work. It did, 
and the payoff was very large, both as a foundation for understanding the problem of 
detection of underground nuclear tests and to seismology as a science. WWNSS arrived at 
a time to have a very great impact also on geology, not in originating, but in confirming and 
extending the ideas of plate tectonics. 

It seems most unlikely tiiat WWNSS, and its consequences, would have existed 
without the ARPA program. On the otfier hand, while responsible for getting it started and 
profiting immensely torn its results, it was difficult for DARPA to continue support for a 
data coUection effort such as WWNSS, even though equipment was updated and the data 
were still useful for nuclear test detection research. The ACDA could have operated 
WWNSS, according to its charter, but was unable or unwilling to do so, lacking funds and 
staff. Congress terminated ARPA fimding for eartiiquake research as irrelevant in 1967, 
tiius forcing a transfer out of ARPA. ITic U.S. Geological Survey (USGS) tiien undertook 
responsibiUty for WWNSS. So far it has been difficult to find die necessary funding for 
WWNSS despite increased interest in earthquake research, at NSF and USGS. 

If and when a nuclear test treaty is initiated, die responsible U.S. agency might be 
involved to some extent in continuing to operate the WWNSS. But the treaty 
responsibiUries would likely involve a network of modem digital seismological instruments 
and computers, linked by satellites, building on DARPA-developed technology, for 
international test monitoring and also for seismological research. 

22 Communication from Dr. R. SprouU. 10/89. 
















SEiSMiC ^ 









^ VELA fc- • * * '1 
AO 104 

h oumzMO 

. J 


HGLP* • • 




SRO* (LP) 















Motivated by the recommendations of the Beikner Panel, a treaty climate indicating 
reliance might have to be placed on long distance detections, the progress in digital data 
processing and some early array experiments,^ ARPA began construction in 1964 of 
LASA, a "large aperture seismic array," an array of subarrays extending over 200 mi. in 
diameter. LASA contained more tiian 500 instruments, with digital outputs transmitted and 
processed on a large scale for the first time using modem telecommunications and 
computing techniques. The construction of LASA was completed in five montiis, in early 
1965, under severe winter conditions. LASA was operated until 1978. 

In 1967 ARPA undertook the cooperative construction, witii die Norwegians, of 
the Norwegian Seismic Array (NORSAR). a "second generation" large array at a location 
outside Oslo. NORSAR commenced full operation in 1971 and is still being used for 
research on detection and discrimination of nuclear explosions. A subarray of NORSAR. 
NORESS, has been outfitted with the most modem seismographs and data handling 
systems and may be regarded as a prototype international seismographic monitoring 



In 1958. the Geneva Conference of Experts suggested that about 170 nuclear test 
detection stations be constructed to monitor compliance with a test ban treaty, die number 

1 E.W. Caipenter, "An Historical Review of Scismomeier Anay Development." Proc. IEEE, VoL, 53. 
Dec. 1965, p. 1816. 

2 -Nuclear Testing Issues," Hearing before the Committee on Foreign Relations. U.S. Senate. 96th 
Congress, 1986. 


and spacing of which were determined mainly by the estimated range of detection of 
possible underground explosions.^ Each such seismic station was to include 
approximately ten short-period vertical seismographs spaced over a few kUometers apd 
interconnected with a recording system by cable. No sophisticated processing wa? 
envisioned. In a 1959 reappraisal stimulated by new data, the Berkner panel on seismic 
improvement stated that some stations should have a hundred or so instruments to bring 
capabilities up to a level approximating that estimated originally by the Geneva experts, and 
that processing and array design could offer potentiaUy great improvements in signal-io- 


Of great importance in the detection and identification problems is die degree 
of Mgnal enhancement that may be gained through instrumental and 
computational operations on the improved sampUng of die seismic data 
made possible by the use of large arrays of seismometers. When the 
operations incorporate die elaborate complex signal enhancement techniques 
that can be performed on special-purpose digital data processmg equipment, 
they may realize an improvement in signal-to-noise ampUmde rano m excess 
of n where n is die number of seismometers in die airay. 

The panel funher recommended die investigation of techniques tiiat had been 
developed for electromagnetic antennas and communications data sampling, and the 
estabUshmcnt of a computer center to move towards die automatic processing of seismic 
data from moiutoring stations. 

In 1959, ARPA set up project VELA Uniform, which began to carry out most of 
die Bcikn« panel recommendations, and about die same time die U.K. began to investigate 
die possibilities of larger arrays. The development in arrays and associated signal, 
processing proceeded rapidly:^ 

Between 1959 and 1963, five array stations were built in die United States 
by die Air Force Technical Applications Center (AFTAQ for die Advanced 
Research Projects Agency (ARPA). which operates die VELA program. 
Each of diese VELA arrays had 10 to 31 elements and 3 km aperture. 
Beginning in 1960. die group under Thirlaway and Whiteway at die United 
Kingdom Atomic Weapons Research Establishment began to urge the use ot 
larger aperture seismic arrays, and built several 21-element arrays ui which 
die elements were arranged in two crossed lines, using vanous apertures up 
to25 knL 

3 Repon of the Panel on Seismic Improvement. Rcf. 1. p. 1 1. 

4 "Experimental LASA Principles," P.E. Green. R.A. Frosch, and C.F. Romncy, Proc IEEE WoL 5X 
Dec 1965. p. 1825. AFTAC. mentioned in this quotadon, had been acuve m seismic de^c^^ 
since 1949. when it was given a naUonal responsibiUty in this area. Early VELA Umform efforts 
depended extensively on AFTAC assistance. 


The U.K. approach was to record broad band, on tape, and use 'Velocity filtering." 
or "delay and sum," of signals from array elements to improve signal to noise. 

In about 1962 the treaty climate worsened, and in the same time frame the Soviet 
and French underground nuclear explosions occurred and were detected at several distant 
scismographic stations, indicating low-loss propagation of compressional P-waves to large 
distances. The UK., followed by the U.S., then began to look into the possibUities of 
detection at large "tcleseismic" ranges (greater than 2000 km), which might not require 
stations in each country, and for "quiet" sites in remote locations where large arrays could 
be installed. 

At Yellowknifc in Canada, a joint Canadian-U.K. 25-km array was built, and the 
Tonto Forest Scismological Observatory (TFSO), in the U.S.. was enlarged to a 10-km 
"Mills Cross" array. Related advances in signal processing were pursued, including 
correlation techniques to exploit signal coherence across the array aperture. At about the 
same time, new developments in small geophones and in low noise amplifiers occurred, 
allowing installations deep in boreholes in an attempt to reduce wind noise. Green et al., 
give some details of these first steps toward larger arrays.^ 

Backus, Berg, and their colleagues at Texas Instruments (T.I.) led in 
developing sophisticated techniques of combining Uie seismometer 
ou^uts into one output 

Acting on the realization that signal coherence over long distances must be 
insured in considering an expansion of array aperture, AFT AC, under the 
initiative of C.F. Romney, set up a network of eight independent mobile 
stations in the TFSO area to form a network having an aperture of 300 km. 
A system of phone line and microwave telemetry leading to a central digital 
multiplexing and recording system was installed in the summer of 1964 by 
AFTAC for Lincoln Laboratory to facilitate data collection and the smdy of 
equipment techniques required for large arrays. 

^ Green, et al., ibid., p. 1826. 


From the experiments with the Geneva-type airays, some information had been 
obtained on noise correlation lengths.^ The instrument spacing in LASA was initially 
smaller than these lengths.^ 

Methods were also worked out to aUeviate some of the anticipated computing 

difficulties of the large arrays:^ 

One of the major criticisms of the large arrays was simply that to use their 
high resolution in an on-line system required tiie provision of many 
simultaneous processed outputs (or beams). This, it was shown» could not 
be achieved witiiout tiiree Stretch computers running in paiallel! Of course, 
no one wants to look at multichannel noise, and in die U. K. work on 
trigger clusters began. These clusters, at die center of each away, would act 
as coherent energy detectors, and provide the "bulletin" data from which the 
choice for subsequent off-line processing could be made. They could also 
provide a trigger pulse to switch on auxiliary processing equipment 
designed to give more detailed on-line analysis. 

In 1963 the first VELA Uniform results were announced and had a strong impact 
on the U.S. negotiations for a comprehensive test ban treaty.^ However, no 
"breakthrough" had occurred. Carpenter summarizes die technical arguments then 
developing for a large array,^^ 

Statistics were accumulating, but of breakthrough for explosions 
identification diere was no sign. Was it time for a new look, a big step 
forward in technology with die hope diat something new would result? The 
eariy doubts about digital computing had been overcome by die introduction 
of special purpose computers, and a whole range of new possibumcs in 
processing were dius opened up. The velocity filtering properties of the 
large arrays, particularly Uieir directional resolution, continued to receive 
attention, particularly since die detection of smaller events increased the 
chances of interfering signals. 

National networics, particularly die Canadian net, and dien die international 
Worldwide Seismic System Networic (WWSSN) were contributmg more to 

6 In his 1971 statement to the Subcommittee on R&D and Radiation of the Joint Commands on 
Atomic Energy, Dr. S J. Lukasik, Director of ARPA, discusses seismic noise coirclauon lengths. 
Hearings bcfwe the Subcommittee on R&D and Radiation of the Joint Committee on Atomic Energy. 
92nd Congress, 1st session, on the status of current technology to identify seismic events as natural or 
man-made. Oct 1971, p. 23. 

7 C J Romney, ibid., p. 90. LASA spacings were eventually increased by decreasing the numbers of 
short-period seismometers. NORESS spacings are smaller, with higher firequency mstruments. 

8 Carpenter, ibid., p. 1720. 

9 Glenn T. Seaborg, Kennedy, Khruschev and the Test Ban, Berkeley, CA; University of California 
Press, 1981. p. 162. 

10 Carpenter, ibid,, p. 1020 


seismology. Regional networks, essentially arrays with fewer 
seismometers but larger spacings than the conventional arrays, telemeterea 
their data to a central recording point. Such networks m Tasmania, 
California. New England, Arizona, and France all began to produce 
seismological data whose value derived largely from the veloaty (including 
azimuth) resolution they could command. 

In trying to elucidate source mechanisms, it was found that geology still 
appeared to be the controlling factor. Perhaps if we could see the signal in 
the microseismic band "the glass would lighten and we would see the 
source less darkly," but this could only come from much larger arrays. 

Then some strange new noise appeared Texas Instruments doing/, k noise 
analysis found si^cant noise power near the origin: high velocity noise. 

On quiet days, Yellowknife showed nothing like the Tn signal-noise 
improvement of noisier days. Here, apparenUy was "manUe P wave noise, 
probably the minimum noise level possible anywhere on earth. Only by 
increasing the array dimension could this noise be effectively reduced: and 
it would have to be a big increase. 

We were also reminded that aftershocks were a feamre of earthquakes. 
Perhaps instead of going to the site of an event we could steer an airay to 
look at it, but again only a large array could provide the required resolution. 

Thus there arose the project for a large array. 

The treaty climate favoring distant observations had persisted, and there was 
increasing appreciation of the large costs dial would be involved in a Geneva-type system 
with 170 monitoring stations. Thus,^^ 

R A. Frosch of ARPA proposed in March 1964 that an effon be made to 
capitalize on existing array art to the extent of actually building a very large 
experimental array. Under his direction, a smdy group was formed to 
oversee such a development 

The "array art" included not only that of radar antennas mentioned by the Berkner Panel, 
but also some of Frosch's own previous experience with construction of large underwater 
arrays, and the associated signal processing.12 Responsibility for LASA construction was 
given to AFTAC. and for the communications and data processing to die Lincoln 

1 1 PJE. Green et aL. p. 1825. and "The Concept of a Large Aperture Seismic Array," by R.A Frosch. 
P.E. Greene. Proc, R. Soc. A, Vol. 290, 1966. p. 368-384. 

12 Discussions with Dr. H. Sonncmann, ARPA. LASA Program Manager, 5/31/88. Frosch P«;jo«s^ 
had been at Hudson Uboraiories in charge of the Navy's Project Ar^^mis, invoWm^^^ 

array, which posed similar processing problems on a smaller scale. Sonnemann. who was also in 
charge of engineering for Artemis, stated that LASA was much less nsky. 


Uboratory.13 Lincoln had participated in some of the early U.S. array experiments, as 
mentioned above, applying digital processing techniques and tiieory based on tiieir 
experience in radar and communications.^^ 

Green, et al.» gives further details about LASA:^^ 

The LASA study group was aided considerably at the beginning by the 
ideas on overall system organization presented to it by the T J. group and 
by the Geotechnical Corporation's comparative evaluations of seismometCTS 
and preamplifiers. An initial rough design was worked out by tiie study 
group, involving 525 sensors and 200 km apermre, and a site m eastern 
Montana was tentatively chosen on the basis of recommendations by T.l., 
together witii noise intensity measurements made earlier in vanous parts of 
die U.S. by die Geotechnical Corporation. This location had many 
desirable properties. It was sparsely populated, relatively uniform 
geologically, remote from oceans, not too distant from known ovCTseas test 
sites, and convenient to transcontinental long-haul microwave faciUttes, 
^ould tiiese be needed. 

In October 1964, T.I. began installing the first two 25-clement, 7-lan 
diameter subarrays, and in December, after it was decided to accelerate the 
program, Teledyne Inc. began installing the remaining 19 subanays, and 
the local telephone companies began open wire line installation. these 
efforts proceeded at a rapid rate in the face of die most severe difficulties 
due to the winter weather. 

"Speedups" oidered by DoD telescoped the originaUy anticipated path of LASA R&D.i* 

System specifications which had been established were altogether 
preliminary and conceived LASA as a huge breadboard which would be 
evaluated in the field on a limited scale prior to installation of die total of 21 
subarray systems. The final design was to evolve from diis step, but much 
experimentation and a considerable amount of systems engmeermg 
remained to be completed. 

A decision by die Department of Defense to accelerate die experimental 
program appreciably foreshortened die operational date. Thus it was diat a 
contract was written on December 1. 1964 requiring full operational status 
on June 1, 1965. 

13 AO 599 of 7/64 for "VELA Large Arrays," and A.0. 624 of 10/64 for "VELA Uniform" to AFTAC; 
A.O. 670 of 2/65 for study of LASA signal processing to the AF ESD (coniracior for Lincoln 

1* "Seismic Discrimination," Final Report, Lincoln Laboratory. 30 Scpu 1982, ESD TR. 82^. 
1^ Green, et al., ibid. 

16 "The LASA Sensing System Design. Installation Operations." C.B. Forbes, et al., /»roc. /£££, Vol. 
53, Dec. 1965, p. 1834. 


Apparently the speedup occurred sometime after Secretary of Defense McNamara 
was briefed in late 1964, and was impressed with the potential for a global test ban 
monitoring system.^ An additional reason for speedup of LASA was to be ready in time 
for the nuclear explosions in Amchitka.i8 aRPA was dien asked to estimate the number 
(and cost) of arrays required for global coverage, which mmed out to be eleven to obtain 
2 to 3 good directional "cuts," at a total cost of several hundred miUions. DoD soon 
decided, however, not to go with eleven but eventually settied for two.!' 

LASA was a state-of-the-art system in its seismic components, many of which had 
been developed under the VELA Unifotm program, and a major step in large-scale real time 
processing. LASA was the first large scismographic system to have digital recording with 
both online and offline data processing.20 There was a new order of magnitude in quantity 
of data flow, and the overall LASA opciaticn was under computer control from a central 
station. Testing and caUbration of the field instrumentation could also be done remotely. 
Fig. 1 shows a "seismic view of the world" from LASA, and Fig. 2 indicates the scale of 
and nature of the installation. Hg. 3 displays a signal flow diagram for LASA. 

The main objective, apparentiy, was to achieve higher signal-to-noise, and obtain 
clearer signals for detailed smdy. 

According to Davies:^^ 

When LASA was being built, it was not known to what extent VN 
(signal/noise) improvement would hold up. The central problem was no 
whether noise would be incoherent at 200 km seismometer scparanon but 
whetiier signal would be coherent over these distances... the an^Y was 
denser in the middle so diat if the signal was coherent only across 50 km, 
more than half the seismometers could contribute. 

While there was considerable argument about what the LASA performance would 
be, when turned on. the majority of statements appear to be that the gain of the array was 
roughly as expected, within a few dB of /N.22 With a randomized distribution of 

n Discussion with Dr. RSonncmann. 6^/88. See also The Advanced Research Projects Agency." 
1958-1974; Richard J. Baibcr Associates, 1975, p. Vn-18. 

18 Discussion with Dr. R. Sproull. 10/89. 

19 H. Sonnemann, ibid. 

20 Digital scismographic recording was pioneered by the oU industry. Cf. Sykes. Ref. 4, p. 246. 

21 "Seismology with Large Arrays." by D, Davies in Reports on Progress in Physics. Vol. 36, 1973. 
pp. 1233-1283. 

22 H. Sonnemann. ibid., see also Lukasik. ibid., p. 29, and P.E. Green et al., ibid. 


instraments, simple delay and sum processing turned out to be about as good as could be 
obtained wiih much more sophisticated processing approaches. 

Real-time beam forming to quickly locate epicenters of seismic sources was 
achieved, and it was possible to issue a daUy worldwide earthquake bulletin. WhUe the 
beams were nairow, the uncertainty of location in the Soviet Union was about 50-100 km, 
too large to be useful for efficient foUow-up inspections.23 In 1967 about a third of the 
LASA instruments were removed, since increased instrument spacings in the subarrays 
reduced short-period noise correlations, and there was no loss of signal-to-noise with delay 
and sum processing.^* No new discriminant between explosion and earthquakes appeared, 
but known discriminants, for sources giving good signal to noise, stood out due to the 
higher degree of signal clarification. The quaUty of LASA signals allowed discovery of 
new reflections of seismic disturbances from inside the earth's core, and also indicated 
large-scale roughness of the core boundary.^ 

Originally it had been planned to construct two large arrays, partly because of the 
need to check one another at what was expected to be a new level of sensitivity, 
unachievable by any other smaller group of instruments. Also, at that time it seemed 
desirable, on the one hand, to have a capabiUty for nuclear detection of tests anywhere on 
the globe, which required use of more than one location to obtain a first "fix," and on die 
other hand to make measurements closer to the Soviet niain test site. Consequentiy, in 
1967 ARPA proposed that another large array, "NORSAR." be constructed, as a 
cooperative project, in Norway. The geology of Uie NORS AR site also appeared to offer 
potential advantages for seismic signal propagation and bandwidth. This array was to be a 
"second generation" LASA, incorporating lessons learned in instrumentation and 
processing as well as automatic detection capaWlity. The instruments removed from LASA 
in 1967 were used to start NORSAR. The Norwegian government approved die 

23 However, when monitoring known nuclear test locauons it was possible to calibrate arrivals and to do 
fine-grained location of new tests on Uie site, H. Sonncmann, ibid. 

2* Early statements, cf. Frosch and Green, ibid., p. 383. indicate early hopes that threshold O^ichter) 
magnitudes of 3-3.3 were expected. However, later statements give a figure of 3.5 to 3.8. Romney, 
ibid., states that the overall gain of LASA was not, in fact, better than that of a smalter anay at a very 
quiet location. 3.9. P.E. Green, a at, discuss the tradeoff of gain and computing costs. 

25 F Ringdal and E.S, Huscbye. "Application of Arrays in the DetecUon, Location and Identification of 
Seismic Events." Bull. SeismoL Soc. of Am., Vol. 72. No, 6, pp. S-201-224, Dec. 1982. 


7 it ■•Ocr< 

Figure 1. The Earth as Seen From the Center of LASA 


Figure 3. Signal Flow Diagram 


project in 1968, and NORSAR became operational in 1971.26 Data links through 
ARPANET connect NORSAR with DARPA's Seisniic Analysis Center in Alexandria, VA. 
NORSAR initially was somewhat smaller, about one-half the size of LASA, and with 
improved understanding of true airay gain, and of computing expense, the array has also 
been gradually "thinned out"^^ 

In 1968 an array of new, low-noise, long-period seismometers was developed 
under VELA Uniform. TTie Alaskan Long Period Array (ALPA), with 19 such instruments 
and an 80-km apermre, was instaUcd near Fairbanks, Alaska. Digitized data was 
transmitted by radio to a local control sensor, and evenmally all the large arrays transmitted 
to the ARPA Seismic Data Center in Alexandria. VA.2* ALPA operated until late 1970. 

NORSAR is still operating. However, Husebyc, et al., state that the large airays 
were in full operation only for about 5 years, during which time a large volume of high 
quality data were accumulated.^ 

In a review article, Husebye and Ringdal state that:^^ 

. the event detection capabUity of arrays has proved superior to that of 
simple stations, but event locations, while readily available, are seldom very 
accurate (not < 50 km) .... the implied two-dimensional wave field 
sampling provided by arrays has been instrumental m understanding 
phenomena like the ambient noise field, die extent of manUe heierogeniety, ^ 
and their effect on short wave propagation. It is somewhat unformnate that 
due to Umitations in handling the enormous amounts of data involved, only 
a relatively small number of seismologists has had access to the high quality 
array recordings; recent advances in computer technology nught elmamate 
such problems in the near future. New technology also makes possible a 
new trend in array seismology, involving on the one hand woridwide 
deployment of small- and medium-sized airays, and on the other hand 
opening up array processing techniques for a global network of such 

interest has shifted more to small and medium high arrays, primarily 
because of cost but also because it has been reaUzed Uiat a few large arrays 
cannot by themselves solve the problems in monitoring a nuclear test ban. 

26 A,0. 1852 of 4/71 for NORSAR computer, to the Air Force Electronic Systems Division. 

27 A review of the status of NORSAR is given in "Seismic Arrays," by E. Husebye and S. Lugati. 
Chapter 28, Arms Control Verification, Boston, MA, Pergamon, 1986. 

28 W£.FaiTcU, -Sensors, Systems and Arrays," in T/ur V£L4 /^rogrom, DARPA 1985, p. 495. Farrell 
gives details of the instruments in the large arrays. 

29 E.S. Husebye, ei al.. "Seismic Arrays for Everyone," in The VELA Program, DARPA 1985. p. 527. 

30 F. Ringdal and E.S. Husebye. ibid. 


Ringdal and Husebye also critically appraise the degree to which the large arrays, 
mainly NORSAR, have been successful in achieving their objectives. A more recent 
appraisal of seismic verification of nuclear testing treaties by the Congressional Office of 
Technology Assessment credits the full NORSAR array, when operated at higher-than- 
usual ftcquencies. with an instantaneous detection threshold of very small explosions in 
selected locations in the Soviet Union 3i Recent results obtained at NORESS, an updated 
dense subairay of NORSAR, at higher-than-ordinary seismic frequencies, have also 
indicated a new possibiUty of detection and identification of smaU explosions, even if 

The same OTA appraisal discusses the relative worth, in cuirent thinking, of arrays 
versus the use of many distributed single seismographs for treaty verification 33 


The LASA initiative was taken by ARPA. The Berkner Committee had made a 
recommendation to look into large arrays. Tht engineering risks taken in tiie expansion of 
seismic array size to LASA dimensions were not regarded as high by die program 
managers. However, tiiere was uncertainty about the results of processing die noise, and 
to what degree signal coherence would be useful across die full apemire. AFTAC, on 
which ARPA had previously reUed heavily, apparentiy did not favor the project, and put 
forward an alternative proposal which ARPA did not regard as involving state-of-the-art 

At the time, it seemed very important to answer the questions of what capabiUty 
could be achieved by pulling together the state of the art in seismic instruments and in 
digital signal processing capabiUty in a really large array. It was envisaged that doing so 
would transform seismology .^^ The treaty climate seemed unfavorable and it appeared that 
monitoring of underground nuclear tests, then considered as possibly occurring in many 
locations on die globe, might have to be done from locations under U.S. control. A very 
large array could give directional indications, and several such arrays were initially 

31 -Seismic Verification of Nuclear Tesung Treaties." Congress of ihe OS. Office of Technology 
Assessment (OTA), USGPO, May 1988. The magnitudes quoicd here are about 2.5. 

32 Ibid., p. 70. Cf.alsoSykes,Rcf.4.p.m TTie bandwidm required for the high fr^^ 

is larger thM can be aaommodaicd by the ARPANET line to NORSAR. Discussion with C J. 
Romney. 7/88. 

33 Ibid., p. 74. 

34 Communications from C. Herzfeld, 1/90. 


discussed to provide localization. Even with several LASA's, however, the locaUzation 
uncertainty was understood to be so large that the problem of follow-on inspection would 
be formidable. 

LASA was successful in demonstrating a new level of data processing capabiUty, 
which has affected all test detection systems since. However, no new "discriminant," for 
nuclear tests versus earthquakes, emerged from the LASA experiments. The increase of 
LASA sensitivity seemed to go as tiie square root of die number of instruments, which was 
less than some had hoped. 

NORSAR, originally thought of as a "second LASA." was closer to die Soviet 
Union, where most explosions of interest were expected to occur. NORSAR was started 
witii instruments taken from LASA. as a result of discussions between ARPA and 
seismologists from Norway, and has been quite successful, indicating the continuing utility 
of the large array concept as a research tool. While no new discriminants were forthcoming 
also from NORSAR at first, recendy die use of high frequencies appear to show some 
promise. NORSAR has also offered a means to assess die cost-effectiveness of smaller 
arrays, of different sizes, and to help define die NORESS subarray. NORESS may be 
regarded as a state-of-the-an monitoring array and a prototype for a international 
monitoring station under a test-ban treaty. 

It is most unlikely that research facilities such as NORSAR and NORESS and dieir 
implications for nuclear test detection systems would exist, without die VELA program. A 
full "transition" of diis DARPA technology has not yet occurred, however, partiy because 
no agency has the ability to cany out an adequate follow-on responsibility. This problem 
may be cleared up if and when a more complete ban on underground nuclear tests comes 
into effcct.3^ 

The ARPA outiay for die LASA facility was apparcndy about $20 million. The 
follow-on research using die facility is estimated to have been about $25 million, for a total 
of $45 million.3« Costs of building NORSAR. including its computer, arc estimated as 
about $8 million.^'^ 

35 "Intelligence Support To Anns Control," Report of the Pennanent Select Committee on Intelligence. 

House of Representatives, USGPO 1987, p. 54. 
3^ Discussions with H. Sonnemann and R. Lacoss, 8/89. 

37 The NORSAR Array and Preliminary Results...." by H. Bungum et al.» Geophys. J.R. Astro Soc. 
(1971) Vol. 25, p. 115 and AO 185Z 













u I 





AO 599 









AO 624 
















7-31 -89-1 M 




ARPA bought a number of lightweight AnnaLite AR-15 rifles under project AGILE 
in 1961 and 1962 to evaluate in VietnaoL The very positive evaluation in August 1962 had 
a major impact on the DoD studies leading to a decision, in early 1963, to purchase AR- 
ISES in quantity for use in Viemam. and eventually on the Army's adoption in 1967 of the 
follow-on M16 as its standard rifle. 


The lightweight, high-velocity .22 caUbcr AR-15 rifle was originaUy developed by 
Eugene Stoncr of ArmaUte division of Faiichild Industries in response to a request in 1957 
by Gen. Wyman of the Continental Army Command.^ The background of this request 
came ftom carUer studies by the Amy's Aberdeen Uboratory going back to the 1920's, 
and in the 1950's by Army supported studies by a contractor, the Operations Research 
Office (ORO), which indicated that a rapid fire, high-velocity, small-caliber weapon could 
be very effective at ranges at which rifles appeared nnost likely, from recent experience in 
Korea, to be used in ground combat^ It was also argued that lighter rifles could allow a 
soldier to carry more ammumtion, and increase combat effectiveness. 

While the AnnaLite AR-15 had undergone a number of tests and had some support 
within the Army, initially it met with opposition from the Army Ordnance Corps. The 
Ordnance Corps favored the heavier, larger caliber. M14, which was designed for use 
primarily in the NATO theatre and had influenced the caHber and choice of and agreement 
on NATO standard ammunition. The semiautomatic M14's were being produced in large 
numbers in the late 1950*s and early 1960's, and were expected to gradually substimte for 

E.C. EzeU, The Great Rifle Controversy, Siackpolc Books. 1984. p. 162. 
Notably, the ORO report "Operational Rcquircmcnis For an Infantry Wwpon." by 1^^^^ 
Hiichmin. June 1952^ see also The Black Rifle, by R. Stevens and E C. ^^^J^l^' 
Publications. Toronto, 1987. p. 9. TTie Viet Cong gave the name of "The Black Rifle to the 14*16. 


several weapons: the Ml rifle, the Browning Automatic Rifle (BAR) and the carbine* as 
these were phased out of the inventory. 

The AR-15 had also been taken on a "World Tour" demonstration in 1959 by 
Mr. Bobby MacDonaid of Cooper MacDonald Company, affiliated widi Fairchild.^ 

In July 1960, an informal demonstration of the AR-15 was given to Gen. Cunis 
LeMay of the Air Force. This led to Gen. LeMay's recommendation for Air Force use of 
the AR-15 to replace their older carbines. After three tries, the Air Force was able to get 
approval for procurement of AR-15's in May 1962.* 

ARP A's project AGILE had a mission of rapid development of material for use by 
Vietnamese forces, and had set up a field R&D unit in Vietnam. The ARPA field unit 
reponed that the small-statured Vietnamese soldiers were having problems with the Ml and 
other weapons they had been given by the U.S. due to weight and recoil.^ Bobby 
MacDonald, now affiliated with Colt Industries, which had bought out rights to the AR-15 
from Fairchild, urged ARP A's project AGILE to test the lighter AR-15 in Vietnam. 
According to Stevens and Ezell:^ 

It wasn't long before the tireless Bobby MacDonald had convinced 
Col. Richard Halleck. on loan to the AGILE team from the Army, that the 
light, lethal but soft-recoiling AR-15 was just the rifle ARPA was looking 
for. By late summer ARPA had officially requested over 4,000 AR-15s to 
support a proposed full-scale test of the AR-15 in conjunction with special 
US advisor-guided units of the South Viemamese Army. This request was 
denied, on the grounds that M2 Carbines were just as suitable for small- 
statured troops, and were available from storage. Undaunted, ARPA boiled 
the whole idea down to what they could afford: a limited range of tests in 
Saigon, in October 1961, with ten Colt AR-15s. The number of rifles 
might have been small, but the enthusiastic reaction of the Viemamese and 
dieir American advisors alike who handled and fired the AR-15s was just as 
Bobby MacDonald had predicted. 

Armed with these positive results, ARPA resubmitted its original rwjuest, 
clearly stating that the AR- 15s required were to be used to aim special US 
advisor units and their Vietnamese allies only, and were not to be 
considered as a general issue item for regular U.S. troops. 

^ Stevens and Ezdl, ibid., p. 83. 
* Stevens and Ezeli; ibid., pp. 87-97. 

5 Richard J. Barber Associates, ARPA History, p. V-44. According to S. Deitchman of IDA ihe equally 
small Vict Cong seemed to have fewer problems wiih captured Mi's. However, R. Sprouil pointed out 
that the differences of operational discipline of the Vict Cong and ARVN also mattered. 
Communication with R. Sprouil 10/89. 

^ Stevens and Ezell, ibid., p. 100. 


This ARPA request came through Military Assistance Advisory Group (MAAG) 
channels. The MAAG had been trying to provide M-l's, which came "firee" as war surplus 
in Vietnam^ In December 1961, Secretary of Defense Robert S. McNamara approved 
purchase of 1000 AR-15's for this field test. ARPA responded quickly, procuring the 
rifles and arranging for shipment.^ The test was to be under combat condidons, and 
involved experienced Vietnamese soldiers and U.S. military advisers. In August 1962, the 
AGILE field test report was in, stating that the Vietnamese much preferred the AR-15's and 
recommending that the AR-15 be considered for adoption by all Viemamese forces, 
especially for jungle combat. Stevens and Ezell, in their recent history of the Ml 6 state that 
"this (report) was ihc most influential yet controversial document so far in the history of the 
already controversial AR-15."^ Because of its interest, most of the field report is 
reproduced in the Annex to this chapter. Immediately after the AGILE field test, the 
MAAG Vietnam requested 20,000 AR-15's. Apparently, the Army Material Command, 
which had absorbed the Ordnance Corps, agreed with the AGILE report that the AR-15 
was more suitable for the small-statured Viemamese troops. However, it was three years 
before AR-I5's were made available in quantity for use in Viemam, and nearly six years 
before they were made available to the Vietnamese forces. 

A follow-on study, by C. Hitch of DoD's Systems Analysis Group, based partly 
on the ARPA field unit study, was issued in late September 1962 and was highly favorable 
to the AR-15. Stevens and Ezell describe the background: 

Over this same period (summer 1962) ARPA staffers back in Washington 
had introduced the ubiquitous Bobby MacDonald to others in the OSD's 
Systems Analysis Directorate. A demonstration for all interested OSD 
personnel was arranged wherein AR-15s and M14s were fired in 
comparison with the standard assault rifle of the communist world, the 
7,62x39mm AK47. Witiiin this framework the AR-15's light weight, low 
recoil and controllability on automatic fire appeared particularly impressive. 

A comprehensive OSD study of tiie history of service rifle caliber reduction 
was soon in the works. Starting with the ,276 Pedersen round of the 
nineteen-twenties, OSD analysts worked their way through the ORO studies 
and BRL's small caliber, high velocity (SCHV) reports of tiie fifties, and 
concluded with the results of their own comparison of the .223 caliber AR- 
15 rifle with tiie M14 and tiie AK-47. A repon of their findings was sent to 

^ R. Sproull, ibid. 

S AO. 298 of 12/61 for AR-IS rifles, project AGILE, to Cooper-Macdonald, Inc. 
^ Stevens and Ezell, ibid., p. 100. 
Stevens and Ezell, ibid. 


Secretary McNamara on September 27, over the signature of OSD s 
ComptroUer. Charles Hitch, Abandoning all pretense that the AR-15 was 
suitable only for small-statured Vietnamese, the Hitch repon stated: 

The study indicates that the AR-15 is decidedly superior in many of the 
factors considered. In none of them is the M14 supenor. The report, 
therefore, concludes that in combat the AR-15 is the supenor weapon. 
Furtheimorc, the available cost data indicate thai it is also a cheaper weapon. 

Although analyzed less thoroughly, the M14 also appears somewhat inferior 
to the Ml rifle of Worid War 2, and decidedly inferior to the Soviet combat 
rifle, the AK-47, which in turn, was derived from the German 
"Sturmgcwehr" of World War 2, 

Because of the contradictory views about the AR-15, the White House requested 
and the Secretary of Defense ordered a reevaluation of the Army's rifle program, to be 
carried out by January 1963. The Army's Chief of Staff had, in fact, already started such 
an evaluation. The Army's January evaluation report was a qualified negative, 
recommending use of the AR-15 for airborne and special forces, but not for NATO. 
However, rumors of bias led the Secretary of the Army Cyrus Vance to request the Army's 
Inspector General (IG) to investigate. The IG leponed a finding of bias. 

After some funher discussion with his systems analysts, who pointed out that an 
Army flechette-firing rifle, the Special-Purpose Individual Weapon (SPIW), was in 
development and might soon supersede the AR-15 and M14's, Secretary of Defense 
McNamara directed in January 1963 that there be no more M14 production after FY 1963, 
noting that there were many M14's in the inventory. The Secretary of Defense also appUed 
M14 production funds to purchase AR-15's for the Army special forces and airborne units. 
The Army assumed procurement responsibiUty for the AR-15 soon after, and agreed to a 
••one-time ' buy of 8,500 AR-15's, which later became 104.000, of which 19,000 were for 
the Air Force. A formal AR-15 project office and interservice technical committee was set 
up by the Army.^i with guidance by Secretary of Defense that changes to the AR-15 were 
to be minimal and at least cost in order to exploit the advantages of commercial 
development. Also there were no RDT&E funds for the AR.15. Deputy Secretary of 
Defense Gilpatrick further advised the Army, "to avoid the cost, delay, and manpower 
difficulties of quality control, parts interchangeable and acceptance test standards programs 

1 1 ApparenUy this was the first technical inierservice committee to be concerned with ^^y^^^^® 
counselled by the Secretary of Defense to consult with Eugene Stoner. developer o the A-IS. about 
any technical changes, but apparenUy this was not done. Stevens and Ezell, ibid., p. i^^. 


of various rifle procurements." However, the Army wanted a number of changes, such 
as manual bolt closure, bore twist, and, importandy, ammunition. The Anny wanted to 
use more potent ball-powder ammunition, apparently in order to obtain larger lethal ranges 
approaching NATO requirements. The Air Force and U.S. Marine Corps disagreed with 
these changes; however, they were instituted, partly because the Secretary of Defense 
insisted on getting a single rifle for all three services, and because of the pressures of 
Viemam. In 1964, the Army type-classified the AR-15 as the experimental M16 EX113 for 
issue to U.S. troops. In the spring of 1965, the M16's were in use by U.S. airborne 
troops deployed in Viemam. In July, Gen. William Westmoreland requested 100,000 
M16's for all American combat troops in Viemam. However, the Commander-in-Chief, 
Pacific (CINCPAC) and the Joint Chiefs of Staff (JCS) disagreed with this request, giving 
as reasons priorities, difficulties with logistics, and the superiority of U.S. weapons in 
Viemam. The intervention of a senator who visited Gen. Westmoreland in December 
1965, cleared the way to satisfy this request.!^ In September 1966, new XM16Ers were 
issued to U.S. Army units in Viemam. In December 1966, Secretary of the Army Resor 
officially informed Secretary of Defense McNamara of the results of the Army's small arms 
weapons systems (SAWS) program, aimed at evaluation of small arms to the 1980's - 
staring that the XM16E1 was generally superior, needed a few further changes, and that the 
SPIW was unlikely to be useful in the foreseeable future, and certainly would not be 
available for Viemam. 

However, as large numbers of M16's began to be used in Viemam, a number of 
serious problems began to be reported, in particular the rifle's tendency to jam under heavy 
use in combat These led to visits to the field by Army and Colt experts, and also to several 
Congressional investigations beginning in early 1967.15 a systematic field test was 
conducted by the JCS' Weapons System Evaluation Group (WSEG) with help from the 
Insumte for Defense Analyses (IDA), to investigate the M16 problems.!^ Some of these 
problems were traceable to a lack of maintenance manuals and instruction, and others were 
evenmally found to be due to excessive chamber pressure associated with the ball-type 
piopellants imposed by the Army, which caused a more rapid firing cycle, and also to 

1 ^ Stevens and Ezell, ibid., p. 125. 

1 3 Ezeil, "The Great Rifle Controversy," p. 180. 

1* Stevens and Ezell, ibid., p. 197. 

15 Hearing of the special committee on the M16 rifle programs (the Ichord hearings) Committee on 

Aimed Services HOR, 90Ui Congress, 1st session. Mar- Aug. 1967. 
1 WSEG Report 164, Operational Reliability Test, M16A1, Rifle System, Feb. 1968. 


corrosion associated with the propellants and the lack of interior plating of the chamber and 
barrel. These problems were considered broadly due to the rapid rate of introduction of 
the rifle directly into use, without concurrent RDT&E, and the corresponding lack of 
proper suppon by industry and the Army, Panly also, some difficulties could be 
associated with the use of more powerful ammunition, in the desire to extend lethal range in 
a weapon originally designed for use at limited range. Some of these problems, e.g., 
maintenance manuals, were dealt with quickly; others have been overcome in a gradual 
"product improvemenL" 

In cariy FY 1968, the M16 was made available to the South Viemamese Army by 
the Secretary of Defense, In July 1968, the U.S. Military Assistance Command, Vietnam 
(USMACV) published an analysis of the results of arming the South Viemamese Army 
army with the M16, which reconfirmed the advantages of size, weight, rate of fire, 
ballistics, and logistics and credited its introduction with a significant improvement of 
operational capability, morale, and esprit de corps?^ 

Many of the problems of the M16 have been gradually overcome by evolutionary 
improvement and change, and the M16 is now the standard rifle for the U.S. Army. The 
The M16 has also been sold, and is in production woridwide. Stevens and Ezell state: 

As sunmied up at an April 1971 ARPA Small Anns Conference by Dr. 
W.C. Pettijohn, author of numerous studies on the analysis of small arms 

The M16 has proven itself to be a superior rifle and has been accepted as 
such on a worldwide basis. It also has potential for mass production in the 
event of an emergency. There are no weapons currently that can be 
considered a competitor. Government efforts to develop a successor will 
proceed slowly. The conference forecasts six to eight million M16 rifles 
being produced during the next ten year period at a cost of two to three 
billion {dollars]. 

Active, direct American military involvement in the Viemam war ended in 
1973. Later Defense Intelligence Agency estimates were that among much 
other ordnance, the U.S. supponed Army of the Republic of Vietnam 

^"^ These corrosion problems had not been noticed in the AR-15, which used a different ammunition, and 
led to statements by the manufacturer that no cleaning was needed for the rifle. This apparcnily was Uie 
reason the M16 had no equipment for cleaning initially, and for statements that no training was 
required. However, the designer did not feel the AR-15 was in ail respects an optimum product 
Discussion with E.C. Ezell. 8^8. 

1 8 "An Evaluation of the Impact of Arming the Vietnamese Army With the M-16 Rifle." Doctrine and 
Analysis Division. USMACV 30 July 68. 

^ 9 Stevens and Ezell, ibid, p. 3 19. 


(ARVN) and the Cambodian Army had been forced to abandon roughly 
946,000 serviceable AR-15, M16, XM16E1 and M16A1 rifles to the 
victorious North Vietnamese Army (NVA), In the mid-1980s, when many 
of these weapons began to appear on the international small arms black 
market, the Ml 6 became the most widely distributed 5.56mm rifle in the 

However, problems remain in meeting NATO requirements for armor penetration 
and also in satisfying requirements of the U.S. Navy with the M16.2° In fact, the U.S. 
adoption of the M16 as its standard rifle appears to have disregarded previous U.S. 
commitments to NATO.^^ Joint Army-Marine Corps efforts were started in the late 1970*s 
under the Joint Services Small Arms Program (JSSAP) program to develop a larger caliber 
rifle and penetrating ammunition for use on future batdefields expected to include large 
numbers of armored vehicles.^ 


The AR-15, predecessor to the Ml 6, was already for sale worldwide and had been 
decided on by the Air Force as a procurement item when ARPA purchased some for test in 
Viemam. Thus ARPA did not undertake a technological development, but a test under field 
conditions which was timely and highly appropriate for the AGILE mission. The train of 
subsequent events, which led finally to acceptance of the M16 by the Army, can be 
definitely traced to the impact of the early ARPA-supponed test results. However, 
ARPA's originally stated motivation, to quickly supply the Vietnamese troops with a 
weapon more suitable for their size and for the short ranges usual to jungle fighting, was 
not achieved. It took nearly six years for the Vietnamese army to get the Ml 6. 

The difficulty in getting Anny acceptance of the AR-15 at the time was partly due to 
the fact that the Army had extensive commitments to the M14, which had just gotten into 
large-scale production, after some difflculties, and had been accepted by NATO, and partly 
to availability of surplus M-1 rifles in Vietnam. Panly, also, ARPA's interventions on 
behalf of the AR-15 aroused considerable resentment in Army circles.^ 

20 "The Great Rifle Controversy," Ezell, p. 250, 259, and 261. 

21 Discussion with S. Deitchman, IDA, 4/89. 

22 Testimony of B. Gen. William H. Fitch, USMC, FY 1980 DoD Authorization Hearings, Committee 
on Anned Services, U.S. Senate. 96th Congress, 1st Session, Part 6, p. 3073. 

23 R. SprouU, ibid. 


The problems with the M16 that occurred in Vietnam can be traced to a mixture of 
DoD overconfidence in the original product, and the changes instituted by the Army 
without concuixent R&D and testing. The lack of R&D was due to a DoD top level 
decision, apparently in the belief that the AR-15 was a finished product, and that R&D 
would get in the way of expeditious procurement 

In spite of die fact that DoD had previously agreed to standards for Ictiial ranges 
with NATO allies, the M16, which does not meet these standards, was adopted as the 
principal U.S. Army rifle. Some of the troublesome changes by the Anny seemed to be 
due to a desire to approach these NATO standards. Apparentiy. NATO may accept 
something like the Ml 6 as a secondary assault rifle. However, expectations continue tiiat 
in a NATO war longer lethal ranges and greater armor-penetrating capabilities will be 
needed, and R&D efforts continue to provide U.S. forces with a suitable rifle. 

ARPA recorded outlay for two purchases of first 10 and later 1000 AR-15 rifles 
and their shipment at a cost of about $500,000. This does not include expense of the 
AGILE field office in Vietnam in connection with the tests. A rough estimate of dollars 
expended for the M16, by die U.S. and others, is between $2 and $3 billion. 













19M ^ MATO 
I 2nd3C 








Advanced Research Projects Agency 
Research and Devdopment FleJd Unit 
31 Mf^ 1962 

Report of Task 13A, Test oiAmaUte RiBe, AR-IS 

The purppucfthu test Mm »deummetf the AR-lSr^ 
ligk ymght cfthe Vumamese Soldier and lo evaluate the 
weapon under aauai combat condinans in Soudi Viemam. 
At the request cfM/^{hBiimyAsasumceMMSor^ 

Vietnam, the scope cf^ test was expanded to include a 
miim^Mdiisatnoresusabie rrpiaoemerttprahershixdder 
weapons in selected ma ef the J<epubUc<^ytetnam Armed 

Farces (FVNAF), 


The prvbUmcfseleaing the most suitable basic weapon 
jbrthe Vietnamese soldier is complicaied by his small suture 
atd light The average soldier stands five feet tall 

ad weighs nineiy pounds. Principle US weapons presently 
issued to ^enamese aoo^ indude die M191842: the 
Thotrpson Ss^Madme Gun, Caliber .45: and the V5 
CaHme. CalAerJO. Ml. 

Because of its avaiiability and the results of extensive 
sadieswd previous testing by military agendes. the Colt 
AmnLUe AR4S r^ wu selected in July, !9S as the most 
suiudtle weapon jbrieidd tests. Thisvmepan\msdevdoped 
by the ArmUte Division cflbirddidAirntfiCorporadon 
to meet the military chamaerisdcs pr a lightweight rifle 
laUizing the high vdodty small caUber principle. Itwasfirst 
tested by the US Army Ir^xnrry Board in 1958. Since then, 
the weapon and its ammunition have undergone extensive 
ai gine et in g ad service tests by: Aberdeen Ptvwig Gwund: 
the Combat Development E\pttiamaman Cam, Ret Ord, 
Ca!^bmia:areithe USAirfbnxatLaddatdAir/brceBase, 
Terns. The rifle, widi several mod^Umans resulting fiom 

iFbwarmsJMaa^aeaeing Con^aty, Hart^, Cameaiaa. 
(Prior to the compledon of this report, the US Air /wee 
adopted the AR-15 as its basic ^hadderMieapan, repladng 
Ote AD Carbine, the Bm/nit^Auumadelbfle and d»eM3 

Sub-Machine Gun). 

fUtmmA uponjitvorabie observations cfthe AR-IS by bodi 
demonstroBons conduaed bi Vtemam during August 1961, 
weapons were reguesud in marben strident to conduct 
afiiU-scale combat evaiiuttionafthe AR-15 by selected units 
cfthe miAF. in December 1961. the Secretary of Defense 
approved the proaament oflSXO AR-lS rifles, necessary 
armmdan, span parts and accessories for evaluation. 

OSD/ARPA negotiated a contnusyMhAe firm cf Cooper- 
Macdanald Inc. , of Baltimore. Maryland, for procurement 
ad air shipment cf dl materiel The first shipment y*as 
received an 27 January 1962 and sutaequeni increments 
arrived apimadmatdy every three weeks laail the contma 
testbtgbegimanl Rinaryadmtnnaedail5AdyJ962. 

The exsremdy mobile type of affensiye warfare being 
stressed by US advisors in Vietnam and the small saaure 
ad ligltt wetg/tf efdie Vietnamese soldier places a hi^i 
prtndian an snuU, Ughtweight weapons. In adtSdan, the 

yioUraAortda^ a dose ranges whkh an chamaerisnc 

cf guerrilla warfare in Vietnam makes it highly desirable 
to have a dependable vieaponoapablecfpTotbicing a hidi 
tnte cf accurate artd iedud fiill ausamadcfin. 


. Mm 

pmm At vie*>poua of standaniaanon ai^ 

buidm.chcmeteristiacfaastingweppcns\mem4aito \y4 
deumine^a^^ weapon cadd be found iha\*x^ 

ntxjps. bisbdievedduttsuchay»eaponshcxM \ 
ihefoUoiitmgdesindjUdiamaerisa^ ^ 

A 7b <lJiiah« range cfthe M7 n>Z& 

Z The tight wei^ and smett site €f die AO Cariiine. 

I Vim automdc eapatUitf cf die BAR, 

4, The simplidty ef the sua 

Odier highly desimble, if not mandatory, faaatrnwuHtd 
inchMk a bayonet, grenade latnching Old adpercapabiSty ^ , 


Details of the Combat Evaluation of the AR-IS 

S e l fr rrd Vtemamese wm yyftieh had prrAousty been 
agaged in eansidefaUe comha werr issued AR-15 rifles 


AR-15 Rifles 


7^ b^antry Diwian 


50W0 rounds 




Airborne Brigade 


I95J0OO muds 

VN Marina 



VN Spedai Btrces 


50000 romds 

Spedai Baaaiions 


J20J0OO rounds 

5th b^mtry A Hsion 


25 WO rounds 

Ridter Hoa 


JOSJOO rounds 



550000 rounds 

Sununaiy of Tests 

cte nw> tMMpoRS wnv compared induded: physiad 
ehamaerisdes; ease cfdisassemiUy and aamUyimarks' 
manship ability at bunm distances, serrd-automaxic and 
auiomaticfire: mariamanship abUity at unhuMn distances, 
senU'OUomaiic ard auumadejire; ruggedness and dum- 
bility; adequacy of safety features: effects of open stiuoge 
in a tropical environment: ability to penetrate dense brush 
and heavy fi)iiage: and, the iiviividual Vtetnanese soUStr 's 
preference betweat the two weapons. 

i/aotmpartx. OneparrMasacombctealuaiiai<ftheARJ5 
in %vhichdte weapons were issued 10 specially selected ARi^ 
units for use in their operations against die Viet Cong. Mong 
widi the rifles and anvnunidon, Viavmese Unit Command- 
ers and l^MditaryAdwon were gpien weapon preferrnce 
and apemdenal questionnaires and requested to complete 

The adier part pf the test eonasied of a camparisen 
bammiheAMSt^taultheMZ Carbine. Anas m¥iadk 

Analysis [of the Combat Bmluatloa} 

Based on die numerical ratings and die comments cf US I b is easy to mabtaan. 

Mvisors and W Urdi Commanders, die AR-15 is die most 

desbobkweaponibrusem Vietnam for the following reasons: 4. bis more rugged and durabU dum present weapons. 

I Bae tfoubung. ^ * bnposes die least logisdeal bunien. 

ZSutOtUpkysiaddiamcterisda. 6 bisOiebestimiponforaB-miduacalen^^^ 


dend aipthor to Out of the Brvming Aiaoataic 

S Vtetnanese troops, Commandm and US Adwon 
pnfiir it to any odur weapon pmtntfy being used in 

Details o/ Comparisoa Jest 

Ikisunn e l fivma l^anmeseeuKpaiy that had jua earn- 

jbrmjscftftisconpansan. 1heiam<]flS)mmym£>Med 
iiaotwo gn)iq»cf90mmeach. Gmip A received ong M2 
Carbineperman, Croup BnscamianAR'JS for eadi 
man. Bach group war dim given a course cfinstrucdtm 
an dieirmpeaive weapon. The inarucdon for each was 
idendcalbidmeaHd scape cfmamoL offered. fbOoMng 

n the AR'IS and M2 Carbine 

das, both groups urtderwaa an ideraiad test prognnnyMdt 
oantisud ct. asamdttf aiul disassembljK AnoMn distance 
firing, bodt send-aaomaacandmtamodejtKe: wdaunm 
isoBKe firing, serid^autonaaietatdauimiade fin: btsyemA 
course; and, ir^Utradon course. This phase lasted fijr one 
^Kek(44houn). Atdtgendc!fdiefirstweek,dietmgnjups 
traded weapons aid diaeourucfitamicdett and die teas 
were repeated. 

Aaalysia [of Coi 

Test I - Physioal Charoaerisdcs 

The AR45 aid die ld2 Carbine art campmMe in size 
Old wm^ and bodt are compadble die light weight 
and small stature cf the VN sokSer. An integiai grenade 
launcher and telescope mount and aiaccesstny bipod are 
included in die weapon weighted the AX-JS. Vmearenoi 
star^ard items fitrthe M2 Carbine. 

Bar 2 ■ Cumpanmve Ease ef Disassembty aid Assemtty 

The AR-15 is simpler and requirta less time lo^aassemble 
and assemble far normal field cleaning. 

Theavemge Vieauanex soldier am be maned in dte dis- 
assembly and assembly fi^ field cleaning of die AR-15 in 
a shorter time than fi>r die iiQ Cabine. This was fitnher 
emphaaud by die fiiadiat ait test subifeas had previously 
received 12 hours (^instruction on die M Carbine mMt 
undergoing basic combat training. 

Test 3' hdarksmaiMp Ain&ty, Knam Disamee 

The abdity of die AR^NsokBer to deliver tteeuaeeeaur 
automatic fire on targets of Ancmfi mnge m«A die AR-tS 
and dot M2 Carbine is comparable. Tat participants, as 
a group, fired a higher penentage of qual^ying scons mdi 
bodt die AR-^ and die M2 Carbine dm diey had pimdaiafy 
fired widtdieM nfiA 

fire on targets of burm range is fitr greater die ASrlS 
rifle dum with die AG Carbine. 

m TestResuUg] 

Test 4 - Marksnanship Ability, Unknosm Distance 

The ABVN xddier's ability to deliver accurate xnu- 
mumantfbeiJiSBig die AlNS and die M2 Carbine is com' 
parable..,die.JibUiiy» deliver accurate M Uo nM i c 
greater widt die AR-J5 dimwididieM2 Carbine. 

"Sat 5 ' Comparative Buggedness and Durability 

iVter die firtt week of firing, seven M2 Carbines were 
eliminated fiom die test. Six of diese would not fire mao- 
maacttllybeeaisecfti^ictPfetSseaimeaar springs; die ciher 
would not fire a all because afa broken disconnects pin. 
In contrast, aUAR-lSsfioKdaned property dumtghaut die 
test period. 

Afier negotiating the Bayonet Assault Course die second 
ame, two AC Carbines wen eliminated fiomdie test beaaise 
if broken stocks. No AR-15 rffies were damaged. 

The AR'15 is considered to be more rugged and durable 
dm die M2 Carbine under condidons which retjuire pro- 
longed firu^ 

The ARiS will staid 1^ ut rough hanMig normally 
encountered in combat situanons better dm die AG 

Test 6' Qmtparison of die Adequacy cf St^iety features 

The st^feamres on die AR'IS and die AG Carbine are 
sokSer's ability to understand them. 


121. A model & Coli AX-lS firm dm lAr order, ierid m OOlty. 

nmo crtda: Ene Umg. Smdaemm /luniwion 

The location of a single selector jmi/c/i w^r/i contbines 
thefiinaions ttf safety selector and rate of fire seleaor. on 
die left side of^ receiver \vhere it is easily accessible to 
the thumb, enables the ARVN soldier to get the first round 
off faster.. than he can with the M2 Carbine. Mth the M2 
Carbine, he must mmipuUae the safety selector h«iA ha 
trigger finger, then return it to the trigger to fire. W/A the 
AR4S he ocn keep his finger an die trigger yMetnmpukang 
the sttfiry seieaor mciA his thumb. 

Test 7 - ^eas of Open Storage in a TnDpical Environmem 

The AR'J5 rifle, because it has fe*ver moving parts, 
will fitnction more readily than the M2 Carbine t^ier 
extended periods of sumge in the open under tropiad 

Test 8 • Brush Peneraion 

The tra^eaory of jru AR-15 buUei is not significantly qfieaed 
t%tienfired through dff-je underbrush a: ranges up to 50 meters. 

The AR-IS round penetrate jungle undergnfi^ equally 
as weU as the M2 Cervine round a ranges 1^ to SO meters. 

Test 9 - Thxip Prenrmee Poll 

The majority ofvu test subjects preferred die AR-15 rifle 
todieM2 Carbine ir. all re^>ects covered by the poll, except 
fiyr the si^tts, Furdir questioning of the subjects by the test 
conmdnee persome. disclosed that this preference was due 
to greaurfiimlianr^ widt Carbine-type agha, na because 
(fminabiiitytowuersuindtheAk-lS sights. This is not 
considered a shoruvming of the weapon but a matter of 
training and familtcissdon. 




It is amdukd that: 

I ThgAMSrijUismconvanbUMthiheUghiweight 
and smaU staatre of the Vietnamese soldier ^ f^ f^ 

Machine Gun. 
2. The AR-15 is superior m the AC Carbine. 

venaiiliTy pr considemiioa as the :Jiic shouider weapon 
for Vietnamese nxxips. 

4. The AR'I5 is capable ''/"P^f.^^^l^ ^i^, 

of the Repabiic of VtemanL 

5. The AR-15 is considered by both Viemamese Com- 
,„andm and US MiiUan Advisors »Wu> parhcipated m 
the tests as the best "aU around" shoulder yyeapon m 


It is recommended diat: 

I The AR-15 be considered for juioption as the basic 
nessandsimpUfimg training and v^vMvUfgisiUs systems. 

2. Ffiorit^'pr adoption of the .^-15 be given u> those 
units ^ch frequentty operate irt . ungie environmem Jor 

^^aive .s^ammunition combinanon a^aiUwie. 

I The Ml and/or sa Carbine continue to be issued only 

tofhoseindiM^ha. ^^^^'^'I'^ '^ZT^r a 
conjunction eSecti^^eh y^th a weapon best smuioiepr a 

defensive role. 

The Project AGILE resuiu from the Vietnam fidd team. 
e.:..piedaLc. summed up b, ARPA back m Washm^- 

;:7. OS lollows: 

The suitabilin- of the AR-15 as tin- Msic shoulder weapon 
for the Viemamese has been estcriished. For the type of 
contUa now occurring in Vietnatr.. the ^veapon ntu idso 
found bv its users and M MAAC u/mo« lo be superior 
in xinuaitvaU respects u} the Ml r7f.e. Ml and Ml Carbina. 
Thanpson Sub-Machine Gun, and BnfMung Autamaac Rifle. 

Test dam derived from recent Serxice evaluanons of the 
ARA5 in the US support the technical conclusions of the 
report. The Central InieiUgence Agency has conduaed 
similar tests: it U understood that the results ojthat 
e^vluaMon are essentiaUy identical to those contained m 
the {above/ report... 




To meet needs in Vietnam, a foliage-penetrating radar capable of automatically 
detecting intruders, named die Camp Sentinel Radar (CSR), was developed by the Lincoln 
Laboratory. CSR was field tested and put into operational use within two years, under 
ARPA sponsorship. The Army copied and improved the radars in a separate follow-on 
program. The processing technique for automatic detection formed the basis for present- 
day commercial acoustic intrusicm alarms. 


In the mid 1960's, camps of U.S. military units in the non-Delta regions of 
Viemam typically were in a clearing surrounded by jungle. Witii limited personnel it was 
difficult to guard against intruders who could come close enough to threaten die camps. A 
need was expressed for some way to automatically detect such intruders in the jungle and 
locate them weU enough to direct fire.i Radar had been suggested as a possible solution, 
but electromagnetic propagation in the dense jungle was recognized as a problem. 

Several programs had been undertaken, with ARPA and Aimy support, to study the 
penetration of jungle foliage by electromagnetic radiation, and a number of related 
measurements had been made in different locations.^ A talk by a DoD representative on 
problems in Viemam sparked interest at the Lincoln Laboratory on the possibilities of a 
foliage-penetrating radar, and their work caught the eye of ARPA staff members.^ Lincoln 
had broad task suppon from the Air Force and ARPA for this and other exploratory work,^ 
Lincoln Laboratory then was encouraged by ARPA to undertake die task of design and 
constroction of a prototype ground-based radar system for test in Viemam.^ 

1 Discussion with S. Dciichman, AGILE Director (1966^), IDA, 10/88. 

2 E.g.. AO 377, of 6/62 for Radar Fbliagc Penetration Research. 

3 Discussion with R. Ziildnd, former ARPA program manager, 7/88. 
^ E.g., AO 498 of 7/63, for Radar Discrimination Studies. 

5 There was also a project to develop an aiibome radar for similar purposes. 


The problem of propagation in the jungle was difficult because of the absorption, 
scattering, and refraction of electromagnetic waves by the foliage, the clutter that would 
result from windblown leaves and tree limbs, and the small and hard-to-distinguish back- 
scattering characteristics of a slow-moving human target near the ground. The radar 
equation applicable to this situation could have several different forms, depending mainly 
on the absorptive and refractive conditions in the jungle, which could affect the design 
parameters of the radar. Using available information on attenuation in the jungle, resulting 
panly from previous ARPA-supponed studies, plus theoretical calculations and 
measurements of absorption by foliage, and scattering characteristics of likely targets and 
of clutter, together with the condition that the radar energy be maximized at a low height 
corresponding to expected targets, estimates were made of polarization, wavelength, height 
for the radar antenna, and required transmitter power. A special analog processing scheme, 
a modification of one previously used by Kalmus of the Army's Harry Diamond 
Laboratories (HDL), was devised to deal with the difficult problem of automatic detection 
of a target having low doppler, wiUiout excessive false alarms, in a time-varying clutter 
environment. To obtain desired rapid scanning, a fixed disc-shaped antenna that scanned 
360 degree electronically with soUd state transmitter elements was also designed.^ Figures 
1 and 2 show a picture of two such antennas.^ 

Lincohi then constructed a fust prototype experimental system which was used in 
extensive tests at CONUS field sites, making measurements of performance, clutter 
characteristics in different types of foUage, and detection of different representative targets. 

In 1968, a second prototype system. Camp Sentinel II, was constructed and sent to 
Viemam for test and evaluation. This second system was almost immediately put to 
operational use at one of the U.S. Division headquarter's camps. Electromagnetic 
penetration losses due to foliage were not as great as had been expected, and good 
automatic detection ranges were achieved. Accuracies were adequate to allow effective 
direction of fire on intruders. Military personnel were trained to operate the radar, which 

K Bowles, et aL, in "Camp SenUnel Radar," J. Defense Research, Sec B., Spring 1969. Vol. IB 
No. 1. p. 66. Unclassified staiements have been made based on this classified article. 
J.R. Dant, in "Camp Senunel Radar III," 18th Annual Tri-Service Radar Symposium Record, Vol. 1. 
pp. 388 and S40. 


Figure 1. Antenna, 30 ft High 


Figure 2. Antenna, 100 ft High 

was moved to another site and again successfully used. Originally, plans had been to 
return the radar to the U.S. after its trials, for modificadons on the basis of lessons learned, 
but because of its success the radar was kept in Vietnam until late in the war, when it was 
sent back to the Army 's Hairy Diamond Laboratories. Laboratory representatives with the 
Army Concept Team in Vietnam (ACTIV) had used the radar in Viemam and had a number 
of suggestions for improvements. Five more Camp Sentinel IH revisions, witii higher 
power transmitters and other improvements, were eventually constructed by HDL and also 


sent to Vietnam. Four of these were used in the field and one for spare parts.^ Lincoln 
wanted to apply a new generic digital processing technique to the Camp Sentinel Radar 
(CSR) in the early 1970s, but instead HDL undenook this task, using the Lincoln 
techniques, and incorporated them into the Camp Sentinel HI radars. The resulting 
completely automatic anti-intrusion radar was used successfully in Vietnam. Two CSR 
m's were returned from Viemam and were used at miUtary installations, and for further 
R&D at HDL. 

The CSR automatic detection processing system was also applied to acoustic 
intrusion detection by one of the Lincoln staff who left the laboratory to form a new 
company. This technique is apparendy in use by most commercial intrusion detectors.^ 


The CSR is an example of a successful, competent Lincoln Laboratory effort, 
undenaken as a result of an ARPA request. CSR was developed and tested in the field 
successfully in two years. Some of the necessary jungle propagation work had already 
been done under ARPA sponsorship to solve an immediate, serious operational problem. 
Perhaps the most difficuh system problem was the automatic clutter rejection, which was 
successfully solved. While all CSR system problems were not completely overcome, a 
successful, workable system resulted, which itself proved so useful operationaUy that die 
original "test'* model was kept in Viemam. This original version of Camp Sentinel was the 
basis for a larger, even more successful, Army program, which was also quickly fielded. 
An IPR was formally issued by the Army, but forgotten after Viemam. The clutter 
rejection technique was also applied successfuUy in commercial acoustic intmsion detection 

In the opinion of some experts. Camp Sentinel, with a new design and highly 
effective performance in the field, was one of the most successful DoD radar 

From project records, about $2 mUlion was spcm by the Lincoln Laboratory effort 
dircctiy on the CSR. Related work on radar penetration of foliage cost about $5 million. 
The benefit was principally in its wartime use. 

8 Discussion with J. Dent, HDL representative in ACTIV Viemam, 12/88. 

9 Discussion with C.E. Muehe of Lincoln Laboratory, 7/88. 

10 Discussion with R. Turner, IDA, 6/88. 



t ,(:; 






AO 491 






/ ' • INTRUSION • • • ' 


I T 






AO 377 

















XVI. THE X-26B-QT.2 


To meet a need in Vietnam for an acoustically stealthy night surveillance aircraft, 
DARPA supported development of the Lockheed X-26B, a powered modification of a 
well-known Schweizer sailplane. While in Viemam, two X-26B's provided real-time 
surveillance as well as test infoiroation for systems improvements. This information led to 
the design and construction of the Army's dedicated, quiet Y0-3A surveillance aircraft, 
which was also used successfully in Viemam. The original X-26B's were given back to 
the Navy test pilot school for use in yaw-roll coupling training. 


In mid 1966 the Army stated a requirement for an acoustically stealthy aircraft for night 
surveillance in South Viemam. Under its Viemam assistance project AGILE, ARPA 
undertook to develop such an aircraft, supporting a proposal by Lockheed for the X-26B, a 
powered modification of a well-known sailplane, the Schweizer SGS 2-32. This sailplane 
was known to be rugged and roomy, and when gliding with power off would be 
acoustically quiet The major modifications included an acoustically insulated and muffled 
Volkswagen air-cooled engine, connected to a large, low-speed, high-efficiency propeller 
by a long line shaft (See Fig. 1), together with an up-to-date sensor suite. Extensive use of 
radar-absorbing paints and other noatcrials was also proposed to reduce radar signature.^ 

To reduce costs and save time, ARPA requisitioned two Schweizer SGS 2-32 
sailplanes which had been recenUy bought by the Navy to give test pilots experience with 
yaw-roll coupling. Witii addition of an observer's seat and some further changes these 

1 Jay Miller, in The X-Planes, Ed., Orion Books. 1987. p. 175. 


Figure 1, The Schweizer Lockheed X-26B-QT-2 
aircraft were modified and designated QT.2PCs.2 The emphasis was on acoustic 
quieting, and reduction of radar signature was not attempted in these aircraft. The two 
aircraft were sent to Vietnam in a C-141 in mid 1968 for a joint-services test under direction' 
of the Army Concept Team in Viemam (ACITV). However, during the Tet offensive the 
QT-2PCS were pressed into service and provided valuable real-time surveillance of enemy 
movements at night After completion of field tests, these aircraft were returned to 
Lockheed for further modification. Two more tours in Vietnam ensued, during which a 
combination of successful surveillance missions and tests to improve capabilities and 
stealthincss were conducted. The results led to design and construction of a new Lockheed 
surveillance aircraft, the Y0-3A. which had new wing sections, new landing gear, a 
modified fuselage., and improved engine and drive system. The sensor technology in the 
Y0-3A was largely determined by lessons learned using the QT-2PCs in Viemam, and the 
Y0-3A mission objectives were virtually identical to tiiose of the earlier aircraft Fully 
dedicated to surveillance. 14 Y0-3A's were built and used successfully in Viemam, and 
only one was lost in action. The rest were returned to the U.S. and used in various ways 
by NASA, border patrols, and the Army. 

The two original QTs were returned to the Navy in 1969. The Navy had bought, 
by this time, two more unpowered Schweizer SQS 2-32 sailplanes (designated X-26A's), 
because of their unique capabilities in training pilots, without undue hazard, iff the 

2 ARPA Order 879 of 4/7/66, "Evaluation of SaUwing Aircraft." and A.O. 944 of 3/67. "QT-2 Low 
Noise Test Aircraft" 


problems of yaw-roll coupling. However, the two powered QTs had advantages of 
avaUabiUty over the X-26A's, since they were able to get into the air under their own 
power. Eventually, one of the QTs was used for spare parts; the other continued in use 
until 1973 at the Navy Patuxent test pilot school It is now in the Army Aviation museum 
in Fort Rucker. 


ARPA's role in the X-26B was clearly the introduction in timely fashion and at low 
cost, working closely with industry, of an effective new combination of available 
technologies ahnost directly into operational use. There was a stated military requirement 
to be met The industry group making the proposal had a very good track record. The 
utility and practicality of acoustically stealtiiy surveillance aircraft was demonstrated and tiie 
sensor packages were tested and proved for use in other programs. An Army dedicated 
surveillance aircraft, the Y0-3A. was designed and produced using the X-26B technology. 

The X-26B-QT-2 apparentiy originated witii a proposal from Lockheed's "skunk 
works." ARPA's role was to work closely witii Lockheed toward meeting a stated military 
requirement, under Viemam pressures. The risks were not very high and lay in tiie rapid 
and effective engineering of a new combination of technologies. An essential move to save 
time and cost was made by ARPA in obtaining existing sail planes from the Navy test pilot 
school. The result was the timely demonstration and operational use of an aeronautically 
stealtiiy aircraft, witii sensor packages tiiat were tested and proved out and used in otiier 
programs at very low cost The original proposal included an effon to make tiie QT-2 
electromagnetically stealtiiy also, but ARPA chose not to do tiiis, probably because it was 
not needed for tiie QT-2's mission. 

Using the X-26B technology, an Army dedicated surveillance aircraft, tiie Y0-3A, 
was designed and produced. The QT-^s powered flight capability was also helpful to tiie 
Navy Test Pilot School and NASA when tiic planes were returned to the U.S. from 
Viemam. The recorded ARPA outiay for tiie QT-2 was $250,000. The benefit was 
principally in its use in VietnaxxL 



























In 1970 ARPA began project POCKET VETO, the first systematic effort to develop 
tethered balloon systems as sensor platforms. Originally intended to carry communication 
relays in Viemam, the concept developed toward combining tethered-balloon platforms 
carrying radar and communications systems with RPVs for surveillance and strike 
missions. Although not developed in time to be used in Viemam, POCKET VETO became 
a joint project with the Air Force, leading to timely deployment, under the SEEK 
SKYHOOK program, of tethered balloons as cost effective MTI radar platforms for 
Southeast CONUS air defense. POCKET VETO technology has also been used in 
commercial TV and communications systems in many other countries, and recently has 
been used by the U.S. Customs Service to begin deployment of a surveillance system for 
the southern U.S. border. 


ARPA effort to develop tethered balloons as elevated sensor platforms goes back to 
1963, with several projects to obtain systems for different altitudes, some as high as the 
100.000 ft altitude range.i Efforts to achieve high-ajtimde balloon platforms continued 
intermittently to the mid 1970s, and the technology developed formed much of the basis of 
the Navy's HASPA developmental program in the late 1970s ^ 

During the Vietnam war, the potential advantages of balloons to elevate sensor and 
communications systems were recognized by ARPA. AvaUable balloon systems were 
procured by the ARPA Advanced Sensors Office (ASO). and tested for utility as carrier 
relays that would assist Army VHFAJHF communications in the jungle. However, these 
first balloons proved fragUe and unstable. Also, the Air Force insisted on limiting balloon 
altitudes in Vietnam to 500 ft. to keep heavily used airspace clear. ASO led an attempt to 

1 Cy. AO's 476 of 5/63 for a High Altitude Tethered Balloon System; and AO 755 and 756 of 8/65 
related research. 

2 AO 2474 of 2/73 NRL: Airborne Tethered Program. 


coiTcct the balloon instability, by aerodynamic analysis, leading to ballasting the tail 
sections. Much of the investigation to correct the instabiUty was associated with the 
concept of using the tethered balloon radar and communications packages, together with 
RPV's, as combined surveillance-strike systems in Viemam.3 Such systems appeared very 
attractive, offering the possibiUty of very low demands on manpower as well as low cost. 

ARPA approached the Lincoln Laboratory to undertake the baUoon-radar project, 
but Lincoln refused on the grounds that the balloon would not prove stable enough as a 
radar platform.^ FeeUng that measured balloon stabiHties were not that unfavorable. ARPA 
ASO proceeded to set up, in 1969, project "EGYPTIAN GOOSE." This project involved 
an available (GFE) Westinghouse Ka-band, aircraft-type, side-looking radar on an 
unstabilized, gravitationally-slung rotational mount hung below some modified barrage- 
balloons, left over from WW H, which ARPA purchased from the UK.5 The radar was not 
fully coherent, and therefore not optimal for MTI, but it was available and could prove the 
concept. Tests were conducted in closed air space in Florida, some of which involved 
tandem balloons to reach higher altitude of about 15000 feet. However, the old barrage- 
type baUoons proved too unstable, and the tandem balloons were difficult to launch. 

Project GRANDVIEW, in the same time frame, involved the same type of balloon 
technology to lift a communications-relay package intended to be used in Viemam. In this 
concept, RPVs such as NTTE GAZELLE, would be able to communicate wide bandwidth 
TV surveillance information, via the GRANDVIEW balloon relays, to ground stations.^ 

The field trials with the EGYPTIAN GOOSE and GRANDVIEW systems had. 
shown both the potential advantages of tethered balloons as intended radar and 
communications platforms and indicated many of the technical characteristics that would be 
desirable for an effective operational system.'' In late 1969 ARPA commenced a project to. 
develop such a system. This program, which took the name POCKET VETO. 

3 -Standoff Sensing," by R. Cesaro and J. Goodwyn, paper at the ARPA ScnsOT Mid Combat Systems 
Symposium. Natl Bureau of Standards. 6-8 June 1970 (Classified). Unclassified excerpts have been 
made from this and other classified references. 

4 Discussion with J. Goodwyn, ARPA POCKET VETO Program Manager. 8/88. 

5 AO's 1521 of 9/69 and 1604 of 3/20. There were 6 baUoons left in the UK, and the Israelivwanted 
some also for similar projects, to enable their electromagnetic systems to look mio Egypt This was 
the origin of the name "Egyptian Goose," J. Goodwyn, ibid. . 

6 The radar used had recenUy lost the compcUtion for radars for a military aircraft system and was 
available as GFE, J. Goodwyn. ibid., AO 1490, 5/69 "EGYPTIAN GOOSE. 

7 "Summary of ARPA, ASO and TTO Programming." Final Report. Vol. 1. B^loons (unclassified), by 
J.H. Brown. M.A. Duffy and R.G. Olilla, Bauelle Report. A65521. Task 44. 1977, p, 22. 


encompassed work in several important technical areas including higher lift/drag 
coefficients, aerodynamic stability in variable winds, materials and structural design, the 
tether and support systems, and safety under various conditions of environmental hazard. 
The program also included development of a MTI radar configured to be used with the 
balloon systems. Several groups were involved in an extensive theoretical work, 
component development, and a field measurements and lest program, notably: the Range 
Measurements Laboratory at Patrick AFB, the NASA Langley Laboratory which undertook 
work on aerodynamic design and test and also on balloon materials; the Air Force balloon 
R&D group at the Air Force Cambridge Research Uboratory on other aspects of the 
baUoon system, including tethers; and, for a time, the Navy Material Command for 
hydrogen gas generators. The NASA Langley Laboratory effort involved constniction of 
model balloon systems for measurements and a number of experiments in wind tunnels.^ 
A 200,000 cu ft balloon was estimated to be required to lift the radar package. Strong 
fabrics originally used in airship construction were tried mitially and rejected as too heavy. 
New materials were developed, with considerable improvements in strength/weight ratio. 
New lightweight power supplies were also designed, simplifying the tether requirements. 
The new balloons, given the collective description of "FamUy 11" (see Fig. 1). were 
subjected to an unprecedented test and measurement program including tow by a helicopter 
at 68 knots to simulate large wind loads. 

In 1972. the Air Force, pushed by Congressional concern stemniing from a 
defecting pUot with his aircraft arriving from Cuba undetected in the Florida and Gulf area, 
conducted several studies of options to meet Air Defense Command (ADC) surveillance 
requirements in those areas. POCKET VETO, by that time, had enough data to allow a 
favorable comparison of its cost and IOC. Although other Air Force groups were opposed, 
AFADC wrote a requirements document for the mission, and in July 1973 ARPA set up a 
joint program with the Air Force for a tethered baUoon platforai to carry a surveillance MTI 
radar for air defense, with a plan for fuU transfer to the Air Force in 1975. The POCKET 
VETO program also involved construction of a new S-band MTI radar designed to have 
improved characteristics for use in the balloon platform. 

AO 1682 of 8/70 Range Measurements Laboratory. "POCKET VETO." Earlier related AO s include 
1666 to AFSC and 1667, both of 5/70, to NASA for "Tethered Balloon System, AO 2176 to RML 
and 2/77 to NASA; 2/78 to NAVMAT and 2/78 and 2/79 to AFCRL ail of 2/72. 



Figure 1. Family II Aerostat Launch 


While POCKET VETO was being pursued ARPA, in response to an approach from 
the Army Security Agency (ASA), set up the joint CEFAR YONDER program.^ CEFAR 
YONDER was to be the first appUcation of the POCKET VETO balloon technology, to take 
place in the NATO theater, with ASA providing the payload to meet field requirements, 
CEFAR YONDER included effon on mobile support systems and a mobile mooring tower, 
together with overall niggedization of the POCKET VETO systems. However, ASA failed 
to get approval for the deployment to NATO. The CEFAR-YONDER equipment was then 
given to the Air Force for the joint ARPA-Air Force project, now named SEEK 

Fonnal transfer of the DARPA project to the Air Force occurred in July 1975. 
SEEK SKY HOOK conducted a successful one-year demonstration experiment in the 
Florida Keys, using a balloon to lift an improved MTI radar for air defense. The SEEK 
SKY HOOK system is now in operational use in the Florida area. Some funher 
developments were undertaken by the Air Force, mainly in the directions of sensor 
improvements and reducing vulnerability to lightning, which has sometime caused the 

The POCKET VETO type of system has also been exploited for commercial use by 
Westinghouse's TECOM division for use as a TV and communications relay in various 
countries. More recently these balloon radar systems, somewhat modified and updated, 
have begun to be used by the U.S. Customs Service for detecting illegal air traffic over the 
U.S. southern border (see Figure 2)M 

POCKET VETO technology is also being studied currently for application to 
CONUS defense against attack by low flying aircraft or missiles. ^2 

9 AO 1876, 9/71, CEFAR YONDER, 

10 E.M. Del Papa and Mary Warner, in "A Historical Chronology of the Electronic Systems Division, 
1947-1986." ESD, Hanscom AFB, Bedford. Mass. 1987, p. 39. Apparently the radars have not been 
damaged directly by lightning, J, Goodwyn, ibid. 

1 1 James Rawles. in^Keeping a Watchful Eye on The Border." in Defense Electronics, Aug. 1988, p. 82. 
and (Fig. 2) USA Today, Dec. 2, 1988. p. 3A. 

1 2 R£, Boisvert, et al., "Tethered Aerostats as Early Warning Platforms," Lincoln Laboratory. Classified 
Report Aug. 1987. 


Customs puts 
new 'picket' 
in drug fence 

By Julie Moms 

The U.S. C'JSiom Service .5 

to lis "picket fence* acifjii 
drug smugglers wiL'i a laur.cii 
Saturday In Demin^ N.M. 

The S18 million ;)iim?.:i« 
ball»K)ns — :fie size of a c:n- 
merc'.al jei — known as aer> 
siais are designed 'o cleiec: zr.d 
deter smugglers along :ne 
USA's southern penmeter. 

The first of six unmar.r.ed 
aerostats that will cover 'Jie 
U.S,-Mexico border from '-".e 
Pacific Ocean to Lhe Cuif of 
Mexico was taunche<l a year 
ago near Sierra Visa, Arii 

"It Is so sophisticaied tf.a: \i 
can monitor traffic on :.1e 
streets of Phoenix" ISO miies 
away, says Charles Conroy. 
spokesman for the VS. Cus- 
toms Service. 

Aerostats weren't always as 
eflicienL Balloons that were 
operating off Florida were 
plagued with radar failures. 

Comparing the aerostats 
vrith the earlier ones is "like 
comparing an F-16 fighter to a 
P-51 World War II fighter." 
says Daniel Wiley of balloon- 

Southern USA radar eyes 

The third U.S. anti-drug radar balloon wil! be launched 
Saturday near Deming, N.IW. By the end of 1990. similar 
aerostats will waxh the entire southern border of ihe USA: 

^ Maximum altitude: 15,000 feet 

above sea level 
> Radar covefage: 32&fliile circle. 

ThB snBoid stanon 

r Size: 23 acres. 
► Crew: 12-16. 


Saurca: USA TODAY rtstarcn 

maker Westinghouse.. 

Supporters say the Arizona 
aerostat is working as a deter 
rent to smuggling. 

"They're sure as hell not go- 

By JuM Siacty. USA TODAY 

ing to fly near it They're driv- 
ing in," says Jamie Ridge, 
spokesman for U.S. Sen. Den- 
nis DeConcini. D-Ariz., Con- 
gress' leader for the project 

Figure 2. Customs Service Radar Balloons 


POCKET VETO was conceived initially because of the need to elevate sensors and 
communications links in Vietnam, in order to operate RPV surveillance and weapon 
systems at longer ranges. It was the first systematic attempt to develop a balloon-radar 
platform system that could meet operational requirements. There were many technology 


risks on an engineering level in POCKET VETO, primarily having to do with stability of 
the platform estimated by Lincoln Laboratory as too difficult to handle, and reliability of the 
overall system. These risks were assessed correctly by ARPA as manageable in a 
determined, scheduled program, and ARPA took the initiative to define and manage the 
program. The technology developments were successful and, while not complete, were 
judged useful as the basis for a military balloon system. The Viemam motivation faded just 
as POCKET VETO was proved approaching completion. Unforeseen Air Force needs 
occurred at the same time, however, and POCKET VETO led quickly to a cost effective 
element in the Air Force air defense system. The direct management by ARPA and the 
close involvement with the Air Force in Viemam-related tests were key factors leading to 
quick and effective transfer of POCKET VETO in spite of opposition on the pan of some 
Air Force groups. POCKET VETO/SEEK SKYHOOK has been in use in SE CONUS air 
defense ever since. 

The POCKET VETO system has also led to a successful commercial venture by 
Wesnnghouse to supply conamunication and TV systems abroad, and to the SOWRBALL 
system, now being deployed to meet cuirent needs for U.S. border surveillance to help 
deal with the drug smuggling problem. 

The cost of development of POCKET VETO, from project records, was about $6.0 
milUon, plus various GFE items that were obtained by ARPA. Predecessor programs 
EGYPTIAN GOOSE and CEFAR YONDER appear to have cost about $3M. For 
comparison, for the new border surveillance system, the acquisition cost appears to be 
about $18 million for a single balloon and ground suppon system. At least six such 
systems are expected to be deployed. 






• • • • • 





9f- . 




The ILLIAC IV, the first array-type computer designed for large-scale parallel 
processing, was constructed with ARPA support in the late 1960s and early 1970s as an 
experimental tool and for eventual operational use on problems requiring intensive 
computation. ILLIAC IV posed a number of major challenges to computer technology 
which caused delays, cost escalation, and reduction in its own size and speed, while having 
at the same time a very significant impact on the general development of computer 
technologies. After reduction to 64 parallel processors, 1/4 of the original number, and 
considerable shakedown, ILLIAC IV achieved operational performance stams in the mid- 
to-laie 1970s, and was installed at NASA's Ames Research Center, under the joint 
DARPA-NASA Institute for Advanced Computation, remaining in use until 1981. ILLIAC 
IV could attain computing speeds in the hundred megaflop range, better than other 
machines available at the time, on several types of important problems for which there were 
algorithms which could be programmed in a way matched to its design. 


The ILLIAC IV was the fourth in a scries of advanced computers developed at the 
University of nimois. beginning with an agreement in 1949 between the University and the 
Amy's Aberdeen Ballistic Research Laboratory.^ The design concept for lUJAC IV, due 
to Daniel Slotnick of the University of Illinois, involved 256 processors in an array of 4 
modules of 64 processors each, under the control of a single instruction unit. A key feature 
of the processor structure was that each processing element could interact direcdy only with 
its nearest neighbor element or the one eight "steps" away. The SIMD (single instruction, 
multiple data stream) concept for paraUel processing used in lUiac IV had originated with 
SOLOMON (a name chosen because it was to have 1000 processors) experimental 
computers, also designed by Slomick and built by Westinghouse in the early 1960s with 

"The Oidvac and the ILLIAC," by James E. Robertson, in A History of Computing inThe 20th 
Century. Ed.. N. Metropolis, ct al.. Academic Press, 1980. p. 34. See also D. Slotnick. The Fastest 
Computer." Scientific American, Vol. 224, p. 76. 


Air Force support.2 This early Air Force effort also included exploration of applications 
and programming of parallel computers.^ 

In 1965 ARPA contacted Slotnick, who had moved to Illinois from Westinghouse, 
and invited him to submit a proposal for a large parallel processor/ Thus commenced 
support of his effort on the ILUAC IV, with the explicit performance objective of design 
and construction of a 256-processor array computer as a experimental tool with a goal of a 
bilUon opcrauons/sec. and with the additional objective of eventual use of the computer on 
various problems requiring intensive computation.^ 

The history of the ILLIAC IV project can be divided roughly into three phases: 
design and construction between 1965 and 1972; installation at NASA's Ames Research 
Center and initial R&D into its utiUty, 1972-1975; and operational use on major computing 
problems, 1975-1981.* ILLIAC IV was formally transferred to NASA Ames by. ARPif^ in 

Between 1966 and 1970 the project was managed by the group under Slotnick at 
the University of Illinois, with Burroughs Company as the overall system contractor, this 
period ended when ARPA decided to have the computer installed not at the University of 
niinois as originally planned, but at the joint ARPA-NASA Institute for Advanced 
Computation at the NASA Ames Laboratory.' 

During the initial design and construction phase a number of major problems arose 
which had both negative and positive aspects. Difficulties with production of chips with 


"The Conception and Development of Parallel Processors - A Personal Memory." by, D.L. Slomik, 
Annals of the History of Computing, Vol. 4. # 1, Jan. 1982; cf. also Paral ei Computers 
Architecture, Programming and Algorithms, by R. Hockncy and C. Jesshopc, Hilger, 1981, p. 16. 
These authors trace the roots of the Solomon computer to a 1958 paper by Unger. ApparcnUy. 
Wesunghouse considered but declined construction of E-LIAC IV which the AEC's Livermore 
Laboratory had planned to lease. ARPA provided all the support for ILUACIV.' 
"Parallel Network Computer. Applications Analysis/ Technical Report RADC-TDR-63-261. 
Aug. 1964. 
Slotnick. ibid. 

ARPA Order # 788 of 10/65,"ParaUel Processing " to AFSC 

The ILLIAC IV, The First Supercomputer, by R. Michael Hord. Computer Science Press 1980, 
pp. 123-132. Page 323-328 of this book gives details of the impact of the ILLIAC IV on computer 

Cf. Sloinik. ibid., and "What Went Wrong V, Reaching for a Gigaflop " by Howard Falk /£££ 
Spectrum Vol. 13, Oct 1976. p. 65. Considerauons of the probability of continumg difficulues at 
the University of Illinois campus (indicated by the riots there in 1970) which could come when:ihe 
ILLIAC IV became operational on mUitary related problems, together with growmg doubts about a 


the desired number of gates using emitter coupled logic (ECL), chosen for speed of 
operation, caused early and drastic changes in the overall design and considerable delay.^ 
On the positive side, ILLIAC IV was the first large-scale user of ECL integrated circuits, 
now found in many high-speed computers. Initially also, thin film memories, based on an 
earlier Burroughs design, were expected to be used, but the changes in design did not 
allow sufficient room. Fortunately. Fairchild had begun semiconductor memory 
development at the time and Slotnick, in spite of criticism about the risk, chose Fairchild to 
make the new memories. The risk in the Fairchild approach involved not only advances 
required in the semiconductor art, but also a number of engineering design and production 
problems. However, Fairchild successfully produced the memory chips, and ILLIAC IV 
was the first large-scale user of these. This intervention by Slomick is credited with 
speeding up the pace with which semiconductor memories, widely used in present-day 
computers, became commercially available.' 

Other serious problems existed with packaging, circuit design and interconnections. 
These posed challenges to the technology which also were eventually overcome, except for 
software, making ILLL\C IV also the earliest successful large- scale test bed for computer 
design automation, now widely used in the industry. Most of the technologies pioneered 
by ILLIAC IV were commercialized within five years-^^ Another novel technology in the 
ILLIAC IV system configuration was a laser-memory system as a tertiary memory with a 
capacity in the trillion-bit range, and read in and out rates in the million bits/sec range. 

These early developments had positive long-run impact on the advance of computer 
technology, but also caused delays and cost escalation for ILLL\C IV.^i As a result, the 

university group's ability to manage such major R&D projects, were some of the reasons stated for the 


8 Initially. 20 (ECL) gates were to be put on a single chip. However, these were not produced 
satisfactorily-leading to a change in design to one using seven gates per chip. A year later, the 
subcontractor was making 20 per chip for commercial use. See Falk, ibid., p. 66. Also, "terminated 
lines" were required, with 60,000 resistors thai had to be changed after deUvcry. Communication from 
P. Schneck, 1/90. 

9 Falk, ibid., p.67. 

10 Falk, ibid., p. 68. 

1 1 The original cost estimate was $8 million for 256 processors. By 1970 $24 million had been spent. 
ARPA set up an independent cost control group in 1971. and by 1972, when installed, the cost of the 
completed computer was $31 million, for 64 processors. For perspective on related costs, the R&D on 
IBM's Stretch Computer, which also stressed the technology of the limc, cost IBM about S25 million 
in 1956-59 dollars, twice the original estimate. See Emerson W. Pugh, "Memories That Shaped An 
Industry." Boston, MA., MIT Press, 1984, p. 183. 


number of processors was cut a factor of 4, to a single module of 64 parallel processors 
instead of the 4 modules with 256 processors originally planned. The processors all "saw" 
the same cable lengths; extra cable was coUed for processors next to the control unit. 
ILLIAC IV's design also provided for a very large main memory and an information 
transfer rate to and from it, involving a novel, accurate synchronous control of discs, 
which could reach the 1.0 gigabit/sec range. Its architecture successfully employed a single 
instruction stream to control the multiple data streams involved in interprocessor 
communication, and used a microprocessor to do this, both significant innovations. This 
1965-1970 period included not only the design and initial construction of the computer, but 
considerable effon on software to exploit the ILLIAC IT s prospective capability.i2 Some 
of the algoritiims developed for the ILLL\C IV. c.g.. "Skewed Storage." are only now 
being exploited extensively. 13 In 1971 Burroughs deUvcred the ILLIAC IV computer to the 
Institute for Advanced Computation OAC) at the NASA Ames Research Center. Figure 1 
is a picture of the installation, and Fig. 2 outiines its design architecture. 

Figure 1. The Computer 

1 2 Falk, ibid., p. 69. ApparenUy this was the first major effort at parallel programming i 

1 3 Discussion with Dr, Paul Schneck, August 1988. 









■ ■ • 







• ■ ■ 



PvalM Organization o4 ILUAC IV tnabttt tti« control unit to oietwttralo %tm 
opontiofl 04 64 precMolng •iomonts, Mch with No own momory. Tlion It a 
lirg* CUM of nuthomatlcal prabiofna that can bo aolvod tn all*ai-oneo 
mannor by Indapondam procaaaora, oporating almuttanaously, 
oaeh about tarteo aa (aat aa tha aingia precaaaer In an 
advancad aaquamlai cempiiiar. 





y — 7 \ 



Block diagram of ILUAC IV tysiam snows how tha ILUACs control unit, 
togathar with tta 64 proeassora and primary mamory untta, will ba 
connoctad to ancillary placaa of aquipmant. A sacondary mamory 
la provldad by a dlak-'flla systom with a capacity of a billion bits 
(binary diglta). A taitlary mamory la provldad by a naw "arehivar 
mamory ayatam, whieti uaas a laaar baam for raading and wrttUifl. 
Aceaasad through a madhim-alza Burroughs B 6500 eomputar. N 
wUl hava atoraga tor a trillkNi bits. 

Figure 2. (from Slotnick, Ibid., Fn. 1) 


During the next phase, roughly 1972-75, the ILLIAC IV was operated by the lAC 
as a R&D project In the period 1973-75 the first experimental "applications" began. The 
ILLIAC was made available eventuaUy to a wide group of users through the Institute for 
Advanced Computation's connection with the ARPANET. ILUAC IV was expected to be 
one of the most important nodes of ARPANET, in order to make its unique capabiUties 
then a large fraction of tiie entire U.S. computing power, available to many users via the 
network. The ILLIAC IV trillion-bit laser memory was an important storage adjunct for 
outside users of the computer, avoiding tiie need to transfer large data volumes on 
ARPANET. Also 10 percent of the laser memory was to serve all ARPANET nodes 
requiring storage, for whatever purpose. However, there were few successes and many 
failures in this period due to the fact that tiic UJJAC IV was not yet operating reUably, and 
because of die real difficulties in progranmiing for paraUel computing. One of die notable 
early successes was on a Monte-Carlo approach to nuclear radiation penetration, for which 
only one of tfjree contractors was able to develop a workable applications program on the 

ILLIAC rv.i* 

In 1975, after a period of intensive effort to correct problems and establish 
reUabUity, the ILLIAC IV was declared operational. Its first use as an operational system 
was for the classified Fixed Mobile Experiment (FME). the first major project of the 
D ARPA Acoustic Research Center established by DARPA's Tactical Technology Office at 
the Ames facility to exploit the ILLIAC IV. FME involved acoustic data transmission by 
satelHte from remote locations, and extensive real time processing. The FME experiment 
demonstrated the feasibUity of the concept as well as the processing capabUity of the 
ILLL\C IV. However, because of reUabUity problems with ILLIAC IV, FME evenmally 
was successfully completed by die Acoustic Research Center and lAC using several PDP- 
lO's in paraUel.!^ 

After die FME. ILLL\C IV became available for routine use. and DARPA directed 
diat die lAC attempt to stimulate die use of die computer for many types of appUcations 
problems. The range of problems dien addressed included, besides acoustic processing for 
die Acoustic Research Center, conq>utational aerodynamics of interest to NASA including 
space shuttie design.^^ several types of seismic problems relating to die DARPA nuclear 
test detection program, atmospheric dynamics, image processing and massive linear 

1* Hold, ibid., p. 124. 

1 5 Discussion with E Smith, former Acoustic Research Center director, 7/88. 
Ifi Business Week, December 6, 1976, "Report on Super Computer," p. 42. 


programming problems. New programming languages were written and a special compiler 
constructed 17 ILUAC IV was itself not a time-shared system, but many users eventuaUy 
had access through ARPANET. Evcnmally most lAC support came from outside the 
original sponsors, and considering that this phase had demonstrated the desired degree of 
utiUty. DARPA turned ILLIAC IV over to NASA in 1979. However, NASA apparently 
did not continue to attempt to obtain a wide range of support, and shut down the ILLIAC 
IV in 1981, but not before a number of design studies had been made at the lAC for a 
follow-on computer, based partly on the ILLIAC IV expcricnce.i^ The ILLIAC IV 
apparently also influenced the Burroughs' BSP computer design, planned for the 
commercial market. BSP was a contender for NASA's National Aeronautics Simulation 
Facility, the follow-on to lAC, but was withdrawn by Burroughs.i^ 

In the early 1970s the ILLIAC IV experience apparently helped Burroughs to win 
the competition to build the PEPE parallel processor for Army's ABMDA, having 
capabilities also in the hundred megaflop range. PEPE was delivered in December 1976 
and apparently met its technical goals almost inimediately thcreafter.^o 


The ILLIAC IV was a pioneer test bed for a number of important advances in 
computer technology, and a unique experimental project. It is widely characterized as a 
failure, along with the other supercomputer designs in die same period, the Texas 
Instruments ASC and CDC STAR. However, the ILLIAC IV was a more radical step in 
design, well ahead of its time, and pushed the technology on many fronts-which led to a 
very high risk of not achieving expectations. In the view of some experts, the failure was 
really of improperly formed expectations, from an experimental projecL^i 

1 AO 2665 of April 1973 for an ILLIAC IV FORTTIAN compiler. Sec also HonJ, ibid. 

1 8 Hockney and Jesshopc, ibid., p. 19. The lACs PHOENIX computer design, for example, is described 
as several ILLIAC IV's under instruction from a central control unit. 

1 9 Hockney and Jesshope, ibid. p. xi, say the withdrawal was due to "producuon difflcuiUes." L. Roberts 
indicates there might also have been uncenainty about commercial markets for the BSP. See Expert 
System and Artificial IntelUgence, by T.C. Bartce. Howard W. Sams, 1988, p. 233. 

20 The System Builders. The History of SDC. by C. Baum. SDC 1981. p. 174. SDC was the prime 
contractor for PEPE. 

21 p. Schneck, ibid. 


Regarding performance, Slotnick has been quoted as stating ihat:^^ 

appUcations have gone just about as I thought they would-no huge new 
computational areas have succumbed to ILUAC. but nothmg we thought 
would work has not worked. 

The performance, of ILLIAC IV overaU, evennially was regarded as better than 
other computers available at the time (see Fig. 3) for several important problems 
programmed to match its structure but far less good for other classes of problems, not so 
well matched. Such a wide spread, often as much as two orders of magnitude in 
performance, remains common to supercomputers.^^ Real-time processing, however, with 
its high demand on reliability, proved difficult to achieve. 

The same advances in computer technology stimulated by ILLIAC IV also caused 
much of its delays and cost escalation. These hardware advances likely would have come 
along somewhat later anyway~but in this rapidly achieving area, time was and is 
considered important. L. Roberts felt that had older, proven hardware technology been 
used in ILLIAC IV there would have been, with some performance trade-offs, a quicker 
and less costly demonstration and evaluation of parallel processing, which was tiie main 
objectivc.24 However, the difficulty of programming for parallel processing was also 
responsible for some of the problems.25 Despite tins difficulty, the ILLL\C IV experience 
apparently "convinced NASA that con^jutational fluid dynamics was a viable alternative to 
the wind tunnel."^ 

22 Hord, ibid., p. 125, gives a sampling of applications problems nm on ILLIAC IV. Besides applied 
problems, HUAC IV was used for fundamental problems in astrophysics and maihemadcal number 


23 See S. Fembach, Appendix A lo The Influence of Computational Ruid Dynamics on Experimental 
Aerospace Facilities," National Academy of Science. 1987, pp. 59 and 71. The performance of Illiac 
IV, accOTding to Hord, was quite close lo what could be originally expected for the 64 processors, in the 
hundred mcgaflop range. However, according to Hockney and Jesshope, the best was in the 50M flop 
range. F6r perspective, the performance of the carUer IBM Stretch over that of the earlier IBM machine 
could vary a factor 100 depending on the problem and the programming. Sec Pug h. ibi d., p. 183. 
Apparendy a similar range of performance estimates s^lied to other "supercomputers" appearing in the 
same epoch as ILLIAC IV. 

2^ L. Roberts, quoted in Falk, loc. ciL 

25 L. Roberts, ibid., and Business Week. Ref. 15, interview with Maicelline Smith of the computer group 
at Ames. 

26 Beyond the Limits - Flight Enters the Computer Age, by Paul E. Ceruzzi, MIT Press, 1989. p. 141. 
However, in the opinion of most aerodynamicists computational fluid dynamics is not so much an 
alternative to wind tunnels as it is a valuable supplement Communication from Dr. A. Flax, IDA. 



Figure 3. Development of Supercomputers • Computer Speed 
Taken from Fembach, ibid., Fn. 7. 


NASA, however, eventuaUy replaced EXIAC IV with commerciaUy available super 
computers. ILLIAC IV did not have a major impact on the next generation of 
supercomputers in the early 1980s. While ILLIAC IV's hardware approach was not 
influential on these super computer developments, it did teach some lessons regarding 
aichitectuic for parallel processors, and in software. 

According to those at lAC closest to the computen^^ 

The ILLIAC IV has taught some important lessons which will have 
significant impact on future parallel processors. In particular, the processor 
interconnection scheme has been found to be wanting. It is both mflexible 
and difficult to program. 

Research in this area has focused on the opdmum interconnection scheme 
and on the most efficient way to use a given interconnection panem. All 
this has been predicated on the assumptions that the connecuon network 
must be fixed (hardwired) and that each processor can be connected to only 
a few other processors (because of fan-out limitations or cost 
considerations). These assumptions are no longer valid since there are other 
alternatives than interconnection schemes based on cabUng, and the next 
generation of array computers should re-focus the attention that the ILLIAC 
has inadvertentiy misdirected. 

Funher, the ILLIAC IV is a fixed configuration with no self-repair 
capability. Current research into self-repairing processors (multi-processors 
such as C, MMP and array processors such as PEPE) are inadequate as a 
base for massive computing power required by scientific computation 
because those prototypes in practice admit only extremely narrow 
bandwidth paths of information flow among processors. Future systeins 
will have modular configurations for improved problem matching and will 
be able to switch aiHng PEs out and good PEs into the configuration all 
under software control. 

TTic chaUcnge of software for a large parallel processor was posed for the first time 
by the ILLIAC IV. and the group at Illinois (Kuck. Lawric, Sameh) pioneered in this area 
of research and education; and made a number of significant contributions which have 
come to fruition only recendy.28 One of the main lessons of ILLIAC IV, apparentiy being 
releamed. is the need to match problem (algoritiim). program and machine structure to 
achieve the highest perf onnance.^^ 

27 Hord, ibid., p. 326. 

28 p. Schneck, ibid. 

29 L. Roberts, ibid. See e.g., "The Synchronous Processor." by Ira. H. Gilbert, The Lincoln Laboratory 
Journal, VoL 1. No. 1. Spring 1988, p. 19. 


The cost to ARPA of the ILLIAC IV itself appears^ from project records, to have 
been about $3 1 miUion. It is widely understood that Burroughs put in $ 1 5M or more of its 
own funds on the ILLIAC development.30 Nearly $28 million was also spent in 
shakedown and utilization of ILLIAC IV. L. Robens, ARPA IPTO director in the early 
1970s, feels that ILLIAC IV more than paid for itself in the cost savings of computer time 
for the problems actually worked out with it.3i An interesting comparison can be made 
with IBM's experience with the STRETCH computer, in the mid 1950s, which also was a 
high-risk project that was expensive for its day ($25 miUion) and did not meet expeaations, 
but had much influence on IBM's later system 360.32 

30 Communication from Dr. P, Schneck, 1/90. 

31 L. Roberts, discussion 7/88. 

32 IBM's Early Computers, by Charles J. Bashe. et al.. MTT Press, 1986, p. 457. 












STARAN .cr ' 














I I 

. « X H « ^4 MMP 
> I 
I 1 

















One of the first major efforts supponed by ARPA's Information Processing 
Techniques Office (IPTO)' was projea MAC^ at the Massachusetts Institute of Technology 
(MIT). In the general direction of broad-based command and control research suggested 
by the Office of the Secretary of Defense, and based on the vision of the first IFTD 
Director, J.C.R. Licklider, MAC was oriented toward achieving a new level of human- 
computer interaction. Within this broad goal, the program included a narrower objective to 
make simultaneous computer access by many users (time sharing) efficient and economical. 
A major outcome of MAC was a large scale, successful effort to develop general purpose 
time sharing, subsequently affecting the design of computer systems for commercial and 
defense uses, generating also many widely used programs for automated engineering 
design, graphics and mathematical manipulation, and grcady facilitating the development of 
Artificial Intelligence (See Chapter XXI). 


1. Early Time Sharing Efforts 

Time sharing of computers for special purposes was not entirely new at the time 
MAC began. SAGE, one of die largest command control systems, constructed in the early 
1950's for air defense, involved some time-sharing feamres allowing multiple access 
on-linc.2 There were a number of commercial systems, e.g., for airline reservations and 

The name of the office subsequenily was changed to the Infomanon Processing Technologies Office 

and then in 1984 to the Informauon Sciences and Technologies Office aSTO). 

MAC stood for both "Machine Aided Cognidon." reflecung the broad research aims of the program, 

and -Multiple Access Computers.- for the actual interactive computer system seen as needed for 

achieving these aims. Sec A Century of Electrical Engineering and Computer Science, by M.W^. 

Wilkes, et al., MIT Press 1985. p. 348. In a time-sharing mode, a computer can be accessed from 

multiple terminals, with several users at once, who have the illusion of Uieirown computer. In 

batch processing, by contrast, a computer is occupied with one job al a ume. 

C. Baum. The System Builders - The Story ofSDC, SDC Corp., 1981 p. 24. WhenAir Defense, 

was eclipsed by the ballistic missUe threat in 1958. the transistorized SAGE computer became surplus 


stock market transactions, which involved some degree of interactive remote multiple 
access to computers.^ There were also some early research efforts at RAND which 
developed time-sharing programs, and at Bolt, Beranek and Newman (BB&N), where 
some programs could be developed and debugged by five simultaneous uscrs.^ C. 
Strachey, of the Cambridge Computer Group in the UK, had given a general description of 
a time-sharing system.^ In the late 1950's, MIT had begun to experiment with time 
sharing using their TX-0 and IBM 704 computers.^ By the early 1960's, in addition to 
MTT, several other university centers also were developing concepts and experiments in 
rime sharing, in particular, Carnegie Instimte of Technology and Dartmouth.'' By 1965 six 
commercial time sharing services had begun.^ 

In the early 1960's MIT had evolved a design for a "Compatible Time-Sharing 
System" (CTSS), working with IBM's Cambridge (user's) group-the first attempt at large 
scale, general purpose time sharing. This system evolved from an experimental system for 
the IBM 709 and first became available in late 1961 using a modified IBM 7090/94.9 This 
was the first demonstration of feasibiUty of a time-sharing system allowing users to write 

and somewhat of a problem to DoD. It was moved to SDC and ARPA was asked to formulate a 
command-comiol program using it This was the beginnmg of IPTO. 

3 Xomputer Time-Sharing: Its Origins and DevelopmcnC by T. James Glauthicr, Computers and 
Automation, October, 1967, p. 23. 

4 Time Sharing Computer Systems, by M.V. Wilkes, Elsevier 1968, pp. 6 and 24. TT^c /OSS ume- 
sharing system, which was developed under ARPA sponsorship, became operauonal at the RAND 
Corporation in May. 1963. Sec Glauthicr. ibid, p. 26. 

5 Quoted in Time Sharing on Computeis," by R.M Fane and FJ. Corbato, Scientific American, Sept 
1966, p. 128. 

6 Wilkes, et al.. ibid., p. 342-343. 

7 Glauthicr. ibid., notes that in 1964 DartmouOi, Carnegie Institute ^^^^^^^^^^^^ 

aU commenced ume-sharing operations. Dartmouth Tunc Shanng System (DTSS) dcveiopmenu 
whichTg^ln 1964 based on General Electric (GE) GE-235 hariwaic. became the basis of GEs 
SiSki commercial time sharing service. Subsequendy. GE and Daronouth col^^boi^ied cm a t^ 
sharing system for GE's 635 computer, which was prototype for MARK n ^^^^^^"^'^^ 
R. Halves and T. Kurtz. "The Dartmouth Time Sharing Neiworic." m N. Abrahamson and F. Kuo. 
Computer-Communication Networks, ftcnticc-HaU. 1973, p. 424. 

8 Glauthicr, ibid. 

9 L. Belady, et al., "Hie IBM History of Memory Management Technology.- IBM Jomial of Research 
and DeZlopment, Vol. 25, No. 5. September 1981, p. 491. Also. Wilkes, et al., ibid, p. 345. 


their own programs. i° Also at this time MIT researchers were developing a time-sharing 
system for a PDP-1 computer donated by the Digital Equipment 

2. Beginnings of MAC 

In 1962, J.C.R, Licklider became the first ARPA IPTO Director. Licklider, who 
had led the time-sharing research effort at BB&N, had a broad vision of the benefits that 
would result to the military and, more generally to society, from progress in interactive 
computing. 12 The corresponding opportunity to undertake a major attack on time sharing 
using the array of capabilities at MIT was recognized by Licklider.i3 in early 1963, Project 
MAC was set up with participation by a wide range of MIT departments, i^ 

The following was the initial research and development program of MAC;i^ 

The broad, long-term objective ... is the evolutionary development of a 
computer system easily and independently accessible to a large number of 
people and truly flexible and responsive to individual needs.... A second 
concomitant objective is the fuller exploitation of computers as aids to 
research and education, through the promotion of closer man-machine 
interaction....The third objective... is the long-range development of national 
man-power assets through education....outside of M.I.T. as well as within 
the confines of the campus. 

The initial MAC time-sharing effort was based on a copy of the latest version of 
CrSS, implemented on another 7094, which was further improved and became operational 
by November 1963. This MAC time-sharing system could accommodate 24 users 
simultaneously. A key role in its development was played by J. McCarthy of the early 
Artificial Intelligence (AI) group at MIT. who recognized the great importance of time 
sharing for the development of AI. 

10 M.V. Wilkes, etal., ibid., p. 342. CTSS was begun on a DEC PDP-l. Glauthier. ibid., p. 25. 

11 Wilkes, ibid., p. 345. 

12 J C R Licklider, "The Early Years: Founding XPTO," in Ej^ert Systems and Artificial Intelligence, 
i.e. Bartce. ed., Howard Sams, 1988, p. 219. Lickiider's vision was initially published as "Man- 
Machine Symbiosis," in the Institute of Radio Engineers Transactions on Human Factors tn 
Electronics, 1960. 

13 Wilkes, ibid., p. 347. According lo Wilkes, Licklider also helped find the first project MAC leader, 
R.M. Fane. 

1* A.0. 433 of 2/63 "Computer Systems," for $8.45M. 

15 R. Fano, "Project MAC." Vol. 12, J. Baker, et al.. cds.. Encyclopedia of Computer Science and 
Technology, 1979, p. 347. 


In the next two years MAC became a general labaratary in which rapid development 
of a wide range of computer programs and techniques took place. One of these, stemming 
largely from the AI group's use of CTSS for symbolic programming, was MACSYMA, 
which has been developed funher into a commercially available package for mathematical 
manipulation and problem solving. Another notable developmem greaUy aided by MAC 
was in the Computer-Aided Design (CAD) area, a graphic display system known as 
KLUDGE. This was an outgrowth of SKETCHPAD, one of the earUest computer 
graphics programs (developed earlier with NSF support), and the MIT mechanical 
engineering department's automatic engineering design effort, also supported by the Air 
Force KLUDGE (see Fig. 1) in turn led to Automatic Engineering Design (AED), the first 
commercial computer graphics program and language.'* SOFTECH was formed by some 
of the developers of AED.'' 

MAC provided a very wide range of "utility" services for compiling, problem 
solving, writing and debugging programs in a number of computer languages. MAC also 
became a large repository for data and programs, raising concerns about losing track of 
content and maintaining some degree of control over access. For reasons like these the 
time-sharing characteristics of CTSS were somewhat restricted in the first two years of 
MAC, while developing a file management system which had the goal of allowing sharing 
without damage, or excessive dupUcarion, with an acceptable level of file security.'* Batch 
processing was also provided for. in "background" or "extra" time." By 1964 MAC 
could accommodate some thirty simultaneous users. 

By this time die limitations of the 7094 for the CTSS had become increasingly 
apparent It had been emphasized in the original MAC research proposal that this computer 
was not adequate as die basis for serious time-sharing system research. The search for a 
more suitable computer started in Fall 1963. and a set of requirements was specified, 

1 6 R. Flamm. Targeting the Computer. Brookings 1987. p. 69. See also Wilkes, et al.. ibid., p. 350- 

l'' Ibid., p. 69. 

' ' R. Fano. ibid. , j j- 

19 The MIT computer center, during all this ume apparently retained its computers mamly dedicated to 
batch processing, as well as the first version of CTSS. Wilkes, ibid. 

20 Fano. ibid., p. 348. 


1 . Read and write protection of user programs 

2 . Privileged instructions inaccessible to user programs 

3 . Direct addressing of at least 250,000 words 

4. A multiprocessing capability with aU processors playing identical roles 
in the system 

5 . An effective telecommunication unit with interfaces to high-data-rale 
graphic display terminals as well as conventional telephone lines 

6. Mass storage units including fast drum for transferring programs in and 
out of core memory 

7 Hardware for efficient paging and segnaentation, including a suitable 
content addressable memory to reduce fetching overhead 

Figure 1. "KLUDGE- Terminal Display 

The "KLUDGE" Display System developed by MIT's Electronic Systems 
Laboratory has a Control Unit Display Screen, light pen and other 

Source: R. Fano and F. Corbato, "Time- Sharing Computers". 
Scientific American, September 1966, p. 130. 


In the words of R. Fano, MTTs Project MAC Director, "It was made abundantly 
clear from the beginning that project MAC was looking for more than just equipment; it 
was looking for a manufacturer sufficiently interested in time-sharing systems to 
collaborate with Project MAC in the development of significant equipment modifications 
and additions to meet Project MACs necds."2i The requirements for paging and 
segmentation were seen as vital, but it was recognized that no commercial computer at the 
time had these capabUities. Widi ARPA approval, these specifications became the basis 
for requested bids from the major computer manufacmrcrs for a new time-sharing 
computer. Proposals from three manufacturers were received: Digital Equipment 
Corporation. General Electric Company, and IBM Corporation. GE won die competition 
with its "635" computer and flexible operating system (GCOS) design, and its agreement to 
be closely involved with MTT in the associated R&D, particularly with regard to additions 
and modifications to meet the last of the requirements (paging and segmentation).22 

In 1965 the BeU Telephone Laboratories (BTL) agreed to join with MAC in the 
development of software (and to acquire the same computer installation), and these two 
were joined shortly after by GE in developing MULTICS (Multiplexed Information and 
Computing Service), and of the corresponding desirable changes of computer design.^ 
A key feature of MULTICS, building upon the original Project MAC specifications, was 
that it would be mainly memory-based with a capabUity to segment and relocate programs 
and data dynamically.^^ 

The loss of this competition resulted in considerable reaction by IBM, as it had been very 
closely involved wiUi MIT's computer activities for many years. IBM had proposed to 
MIT the development of a multicomputer modification of its 360 series, incorporating some 
additional time-sharing features. However, these apparentiy lacked flexibility, specifically 
the feature of "dynamical relocation" of programs in and out of core memory 

21 Ibid 

22 Ibid. MAC also purehased a PDP-6 as a peripheral processor. See Franklin M. Fisher, ct al., IBM and 

the US. Data Processing Industry, Praeger 1983, p. 160. 

23 Fano. ibid., and Wilkes, el al., ibid., p. 351. 

24 Fano ibid p 349. J. McCarthy, who left MAC in 1962, had ouUined most of these requirements in 
196l! The Alias Computer at Manchester. UK had pioneered some of the desired memory orgamzauon 


specified by MAC ^5 EBM apparently had done some work on time sharing but their 
market analysis indicated exploitation of the other features of their 360 series would be 
more important commercially .26 Shortiy after losing this competition, IBM supplied a 360- 
based time-sharing system to the Lincoln Laboratory, which IBM regarded as 
experimental, and in early 1965 began to work closely with Lincoln and several other 
leaders in the field on a broad research effort in time sharing. The IBM R&D work by this 
time was considered by some of the MAC leaders as comparable in scope to their own 
efforts on MULTICS.27 ibm persisted and in the 370 series in the early 1970's marketed 
a time sharing and "virtual memory" system, with architecmre differing from MULTICS 28 
The MULTICS effort at MIT and GE lasted about five years and proved to be 
considerably more difficult and costly (a factor of two) than originally expected. It was 
impossible to "simulate" such a new experimental system and several design iterations were 
found to be necessary before MULTICS could be available for general use in 1969. By 
1971, MULTICS had some 10^ words of procedure code, and served 55 simultaneous 
users, 22 hours a day, 7 days a week, with only one or two "crashes" in a day.29 
MULTICS incorporated a number of very advanced features: a modular structure 
decoupling physical storage and files organization,30 "vinual memory" and dynamic 
reconfiguratton-notably into operating and developmental subsystems, which could be 
done routinely 5 to 10 times a day. MULTICS included an automatically managed 
multilevel memory, and had multilayer supervision of procedures for protecting 
information. MULTICS used the programming language PL-1, which was available at the 
time, and was able to accommodate many other working languages. A very popular 
feature of MULTICS was that, once logged in, a user or sets of users could have their own 

25 Fisher ibid, p. 160-7. discusses this reaction in some detaU. IBM had actually been working on the 
dynaniic relocation capabiUty but did not include it in their proposal to MTT. See also "The System 
360, A Retrospective View," by Bob D. Evans, Annals of the History of Computing, Vol. 8. No. 2, 
1986. p. 171. 

26 Evans, ibid., p. 175. 

27 "MULTICS-The First Seven Years," by FJ. Corbato, ct al., AFIPS Conference Proceedings. Vol. 40. 
1972, p. 572. 

28 "The Origin of the VM/370 Time-Sharing System." by R.J. Creasy, ^^^/v i4iSr'/ZrlVfn^ 
Development, Vol. 25, SepL 1981. p. 483. Evans, ibid., shows the rapid growth of IBM's market for 
time sharing and networking computer systems, greater than IBM had expected. 

29 Corbato. ibid., p. 571. 

30 Ibid., p, 573. 


apparenUy "closed" subsystem. The structure is indicated in Figs. 2 and 3. By 1972 
MULTICS had become a useful and flexible general purpose computer utility and while 
stiU evolving to some extent, was judged mature and turned over to the MIT Information 
Processing Center.^i 

Honeywell, which had bought out GE, supplied the modified 635 computer, now 
called a 636, to MAC for MULTICS, and by the time of its transfer to the MTT Information 
Processing Center was to further supply a "6080," internally nearly identical to the 635, 
The 6080 type, together with software derived from MULTICS, was then being sold 
commercially by Honeywell, Over eighty of these computers were eventually bought by 
miUtary groups, e,g„ Air Force (RADC and Air Force Data Centers) and by the World- 
wide Military Conunand and Control System (WWMCCS) in DoD and its field stations.32 
Later, efforts continued in several places on multilevel security aspects of MULTICS, and 
on other applications including image processing and Computer Aided Instruction (CAI).^^ 
However, "retrofit" MULTICS security modifications offered by Honeywell were not 
bought by WWMCCS and DCA, because of cost and certification problems.^* 

By 1969 die major goals of MAC were felt to have been achieved .35 MAC became 
one of the main nodes of the ARPANET in 1970, and continued for several years as a 
research project on such topics as robotics and automatic programming. The AI group 
working with MAC had grown and in 1971 became a separate laboratory. In 1975 MAC 
ended as a multidisciplinary project and further research activities were continued at MTT 
under the Laboratory for Computer Sciences. In 1987 MULTtCS was shut down at MTT. 

3. Other Developments in Time Sharing Systems 

In 1969 the BTL group involved with MULTICS returned to their parent 
laboratory. Shortly afterwards key members of this group, reacting to their MULTICS 

31 Ibid,, p. 580. 

32 See testimony of G. Dinecn, Hearing before Defense Subcommittee of the Committee on 
Appropriations, H.OJt., 96th Congress. 1st Session, p. 248 ff, 1979. 

33 "Evaluation of TICS," a MULTICS Subsystem for Development and Use of OAI Course with 
MTTRE, ESD 75-76, 1975. Also J. McCarthy had gone to Stanford from MTT and in 1963 designed a 
time-sharing system for experiments conducted there by P. Suppes. Discussion with D. Fletcher, IDA. 

34 Discussion with Dr. I. Bialck. JCS, 3/89. See also. "MULTICS Security Kernel Validations. Vol. 1" 
by Ames, ed.. MITRE. ESD TR-78/48 July 1978. MULTICS was considered the first control system 
designed from the beginning with security in mind; one of its motifs was to protect MTT users from 
mischief and plagiarism. 














— ( [ CARD- PUNCH 


^ '"} PRINTER 










Figure 2. Simplified Schematic Diagram of Principal Elements of the MIT Time- 

Sharing Computer Installation 

Source: R. Fano and F. Corbato. "Time-Sharing Computers", 
Scient^c American, Scptemljer 1966. p. 135. 

35 See Fano, ibid., p. 352, and discussion with Dr. I. Bialek, 3/89. 


'i ;:lnerator 



m. U'm •«•« tlM rmnl <mH wkk mUik mht aau, 
I .iMiti davinf arc m tSun • MNwImt aman- 




Figure 3. "Users View" of the System Is Quite Different 

Each of the 30 On-Line Users has available for all practical purposes, his 
own processor and memory. Each memory has in effect a capacity of 
32.768 words and has access to public files as well as the user's own files. 

Source: R. Fano and F. Corbato, "Time-Sharing Computers". 
Scientific American, September 1966. p. 136. 


experience, invented UNIX, a simpler system allowing the type of flexible, cooperative 
remote computer usage that seemed more appropriate for professionals at After 
some successful experience intemaUy at BTL, UNIX has become available commerciaUy 
and is in widespread use largely in a DARPA-supported modification by the University of 
California, Berkeley.^' 

Another major early time-sharing R&D effort supported by ARPA was at Systems 
Development Corporation (SDC).38 The Q32 computer initially designed as a 
transistorized upgrade to the SAGE system was given to SDC to be used for the ARPA 
command-control R&D program, SDC had been a key participant in several command- 
control system designs, notably those of the Air Force "L" systems. However, the SDC 
work was redirected to emphasize time sharing by Licklider when he became first IPTO 
director in late 1962. This redirection included a demand for a working time-sharing 
system, based on the Q-32, in six months. This was accomplished by the experienced 
programming team at SDC and the resulting time-sharing system (TSS) design won the 
AFIPS prize the following year. This SDC Q-32 TSS was linked by teletype with MTTs 
CTSS and demonstrated at MAC'S initial summer smdy, in 1963. 

The SDC TSS, together with advanced display systems and a more flexible 
language, evolved into a new time-shared data management systetxi, TDMS, leading in mm 
to ADEPT, which accepted nearly namral-language computer commands and which could 
be operated initially on the time-sharing IBM 360/67's and later on other computers. 
ADEPT incorporated special provisions for security, and beginning in 1968 was used for 
some time at the National Command Center (NCC) and the Air Force Command Center. 
SAC also used ADEPT for its sums reporting system, for which it later took back the Q-32 
computer from SDC to SAC HQ at Omaha.39 ADEPT also was the basis for the TIPI 
tactical information processing system, designed for the Air Force in 1968 and entering 
procurement in 1971.^*0 The TDMS, in turn, while suffering some early business- 
application oriented setbacks, led to further applications such as. MEDLARS and the 

3 The name "UNIX" was to be contrasted to MULTICS-to emphasize the coopeiative, as opposed to 

proprietary features of program generations associated with MULTICS. 
37 "A Short History of UNIX," Electronics, March 14, 1981, p. 126, and "Evolution of the UhOX 

Operating System," ibid., July 28, 1983, p. 115. 
3B Baum, ibid., p. 91. 

39 Baum, ibid., p. 1 19. ADEPT was eventually abandoned by the NCC, however, due to slowness in 
turnaround. Discussion with N. Jorstad. IDA, 2/89. 

40 Ibid., pp. 123 and 171. 


associated medical information retrieved system MEDLINE, and later to SDCs own 
commercial information retrieval service."^^ 

TOPS 20, the DEC Company's Commercial Time Sharing Systems, was also 
impacted by DARPA supporting the TENEX operating system.^2 


MAC was an ARPA initiative, part of the broad vision of the first IPTO director, 
LickUder. who focussed on general puipose "time-sharing" as die next major development 
to make computers more useful. There were internal obstacles in that the ARPA director. 
Roben Sproull. was not entiiusiastic at first, feeling that computer development should be 
left to companies iilce IBM. After a visit to several laboratories witii UckUder. however, 
SprouU became convinced that IBM was mainly interested in large-scale commercial batch 
processing applications, and not the technology needed for time sharing and command 
control problems and that ARPA should do something to develop this technology.^3 

Rather than attack the command control application head-on Licklider felt that a 
research effort to develop the broad capabiUties needed in the long run would prove more 
useful.^ MIT was an ideal acadenuc environment for MAC, already having a large 
number of participants stimulated by the earUer CTSS development, such as die strong 
groups active in engineering graphics and AI and recognizing that a big step beyond CTSS 
was needed. Not only was this next development, project MAC sponsored by ARPA at 
MIT, ARPA also played an important role in sponsoring several other time-sharing systems 
in the first years. "In fact, of tiie first twelve systems developed, ARPA participated in the 
sponsorship of six of rhcm."^^ The early contributions fiom the AI group at MTT were very 
significant; time sharing was realized (before MAC) by J. McCarthy of that group to be an 
essential tool for rapid progress in AI. Time sharing was also understood to be very 
important for Computer Aided Instruction. 

Perhaps the main national impetus towards time-sharing development had been 
accompUshed by 1965, with commercial systems springing up at several places and 

^1 Ibid., p. 183. 

^2 Flamm, ibid., p. 58. 

43 Discussion with Dr. R. Sproull, 3/88. 

R. Sproull. ibid. 
45 Glauthier, ibid., p. 25. 


commercial services beginning to be sold about that time. While some of these seem to 
have grown independently of ARPA and MAC it also seems clear that nothing like the rate 
of progress in the area would have existed without the ARPA support for MAC. The next 
step beyond time sharing, computer networking, also a pan of Licklider's early vision, 
soon began to develop, stimulated by the success of MAC and other time-sharing efforts, 
while MAC was still going on. 

The MULTICS initiative seems to have been MTTs, as a natural "second generation 
time-sharing" effort As a cooperative software-hardware effon it was one of die very few 
of this kind. MULTICS led to development of some hardware features of the Honeywell 
6000 computer series, and direcdy to the associated software. MIT has a tradition of 
effective "technology transfer" to indusory, illustrated in this case by working together first 
with IBM for the CTSS, and later with GE and Hfoneywell . Their time-sharing capabilities 
and the desirable features of the GCOS operating system were key reasons why die GE 
computers were selected by MAC.^ The Honeywell 6000-series computers seem to have 
been a fairly successful commercial product, and were widely used by DoD. 

MTT's selection of GE for MULTICS seems to have caused IBM to move much 
more rapidly toward time sharing than otherwise, and thus had considerable commercial 
impact. While MULTICS and the 6000 series were delayed due to underestimation of 
difficulties in achieving time sharing capabilities with acceptable level of flexibility and 
security, much the same seems to have happened in the later IBM time sharing effort. A 
positive result of MULTICS delays and problems was in the reaction of the BTL 
participants, who went home and invented the simpler UNIX system, partly as a reaction to 
MULTICS' characteristics for protection of information, desirable in the university and 
military environments, but which somewhat inhibited cooperative work by professionals at 

By the early 1970*s time sharing had become the dominant mode of computer 
operation in military, business, and academic centers. About the same time as IBM's 
introduction of its VM-based systems, DEC*s mainframe computers adopted time sharing 
as an integral aspect of their systems. Subsequent developments in microelectronics 
technology, both in memories and logic devices created the personal computer (PC) and 
specialized work stations as alternatives to time-shared mainframes. While the rapid spread 

GE's operating system, GE COS, was considered the best at the time and influenced IBM's 
development considerably. Discussion with W. Mulroney, IDA, 2/89. 


of PCs and work stations has, to some degree, overshadowed the time-shared mainframe, 
the advent of supercomputers has further stimulated time sharing locally and remotely via 
networking. The interplay of these technologies continues as technical and economic 
factors drive solutions to computer systems. 

The MULTICS-based approach toward multilevel security was followed up in 
R&D by the Air Force, but not picked up by the DoD, apparentiy due to concerns primarily 
regarding certification and related cost .^^ 

ARPA expenditures for MAC are estimated from MIT records as about $25M for 
die 1963-70 period.^ The WWMCCS had spent, by 1979. about $700M on Honeywell 
6000-typc computers, peripherals and software.*' By the mid 1970's nearly every 
mainframe computer sold had time-sharing capabilities. 

^"^ N. Jorstad» ibid. 

Report on Sponsored Research, MIT Archives. 
49 Hearing Department of Defense Appropriations for 1980. 96th Congress, 1st Session 

Testimony of Dr. Dickens, p. 248. 





ARPA effon on packet-switching technology to achieve efficient, low cost 
intercomputer communications was initiated by Lawrence G. Roberts in 1967, Unking 
selected IPTOi contractors. In 1969 ARPANET, the first wide area general purpose packet 
switching computer-communications network, was set up. Unking different types of 
computers over leased communications lines. Evolving as an experimental network, 
ARPANET operated for several years with scientific measurements and analysis results 
openly pubUshed, and was soon extended to include experiments with packet speech, and 
with radio and satellite communications links. From the early 1970's ARPANET 
technology has been used to an increasing degree in successive generations of DoD's data 
networks. ARPANET also led directly to TELENET, the first U.S. commercial pubUc 
packet switching communications service, and its technology has been the basis of most of 
the many worldwide commercial and common-carrier data networks. As these networks 
grew and required interconnections, ARPANET software research and experience has 
provided much of the basis for network intercommunication protocols. With the increasing 
need for wider bandwidth networks, ARPANET will be replaced by a Defense Research 
Networic. incorporating a new generation of packet- switching technology. 


ARPANETs history can be divided into several phases: (1) a gestation and 
planning phase from mid 1960's to about 1969; (2) an early development and 
experimentation phase, from about 1969 to 1972, culminating in a significant public 
demonstration in 1972; (3) an initial implementation phase, from about 1972 to 1975. and 
(4) a DoD-widc implementation and commercialization phase from 1975 onward. 
Significantly, the 'Ttefense Data Network" (DDN) for interactive communications is based 
directly on ARPANET technology. Research on the extension of ARPANET packet 
switching technology into other media and applications also has been conducted from the 

1 ARPA'sInformauon Processing Technology Office. 


early-1970s. With the prospect of a national research network requiring much wider 
bandwidths, current plans are that the ARPANET will be replaced by a "defense research 
network" more mned to new capabilities. 

1. Origins 

J.C.R. Ucklider, the first ARPA IPTO director, had a vision and a broad program 
for developing man-computer interaction technology.^ After time sharing had been 
demonstrated and its impact began to be widespread in the mid 1960's, the next logical step 
in this program was the linking of computers and terminals by communications networks, 
so that computer capabUities, programs and file resources could be accessed readily and 
shared remotely. The mainstream of ARPAOTT development involved individuals and 
institutions in the computer research communities which were supponcd by the growing 
ARPA IPTO program. However, related early work was done by others, including several 
private networks and laboratories. 

Notable early contributions had been made by P. Baran and coUaborators at RAND. 
Baran's work in the early 1960's outiined a distributed, survivable digital communications, 
system for the Air Force, in which a data stream would be broken near the ppint of 
initiation into addressed sub-units of less than two hundred bits, which would then be. 
routed by "intelligent" nodes over multiple paths which could include satellites as well as 
telephone communication lines. Baran's group also ran a simpUfied computer simulation 
of such a network, using six nodes, which demonstrated its woricabiUty and survivability 
and indicated that the nodes did not need to store many message segments in order to be 
effective.3 Baran's work also showed that such a distributed system would b$ -more 
economical tiian conventional communication for "bursty" data exchanged by a sufficiendy 
large number of computers.^ A 1962 thesis by L. Kleinrock. then at Lincoln Laboratory, 
came to a similar conclusion. The Air Force did not follow up Baran's work, apparentiy 
because of skepticism from the communications conomunity. which fell that data hang-ups 

-Man Computer Symbiosis," by J.CA. Ucklider. IRE Trans, Human Factors in EUctromcs. Vol. 1. 
I960, p, 4. . _ 

-On Distributed Communicauons Networks." by P. Baran. IEEE Trans on <^?'l^^["';'J^^^^ 
wSch 19S ApparenUy Baran's woric at Rand dated back at least to 1960. cf. A. Wohlste^r and R. 
S^y "^ntinu^Con'trol as a Requirement for Deterring." in A. Carter et ai.. eds.. Managing 
Nuclear Operations, The Brookings Institution. Washington. D.C., 1987, p. 1 /a. , - 

L. Roberts,"The Evolution of Packet Switching," in ^^^^^^ 
Distributed Telecommunications: A Compendium of Source Materials, Lifeume Learning 

Publications, 1984. p.Ul. 


would be common and buffer storage requirements large ^ Baran's work. apparenUy, was 
not well known to members of the DARPA community when they began their plans for 
computer communications networks. 

In 1965. D. Davies of the UK's National Physics Laboratory (NPL) gave a seminar 
at MTTs ARPA-sponsored project MAC (sec Chapter XDQ in which he oudined several 
ideas about what he later named a "packet switching" network. Returning to the UK, 
Davies proposed such a system to the British Post Office, which expressed interest but 
responded slowly. Davies also set up a minimal prototype packet-switching netwoik at 

One of those at Davies' MU seminar was Lawrence Roberts of Lincoln Laboratory, 
who had by that time been involved in experiments (also supported by ARPA) carried out 
at Computer Corporation of America (CCA), linking the Lincoln time-sharing TX computer 
with the SDCs Q32.S This experiment indicated problems because of the slow switching 
times of the telephone diaUng system and the noise of telephone lines designed for the 
relatively long and "forgiving" nature of voice communications. Robens recounts that 
earlier, on the basis of discussions with LickUder and others at a meeting in 1964, he had 
concluded that time sharing was launched and that die next important step was to design 
computer-communication links from the computer point of view.? Alternatives to special 
intercomputer communications systems, such as developing a ''universal language" for all 
computers, or demanding aU computers be designed to be compatible with 
communications, seemed impractical. 

At about the same time there had also been a number of inter-computer links, as an 
outgrowth of time-sharing at other laboratories, in industry, and academic institutions, 
notably the OCTOPUS system at the Lawrence Livermore Laboratory linking large 
computers^, experiments at Bell Telephone Laboratory (BTL) on load-leveUing by linking 
similar computers, and in the SITA airline reservation system. OCTOPUS apparenUy 
used a technique similar to packet switching, but did not give the technique a name.^ The 


L. Roberts, unpublished address, 1985. 
6 "Toward a Cooperauvc Network of Time-Shared Compuiers." by T. Marill and L. Roberts, Proc. First 
Joint Computer Congress, 1966, p. 425. An carUcr time-sharing link of these computers had been 
demonstrated in project MACs first summer study. 
L. Roberts, ibid. 

8 D Pehrson, "Interfacing and Data Concentrauon," Chapter 6 in Computer-Commmication Networks. 
n! Ahrahamson and F. Kuo. eds. Prentice-HaU. 1973. describes the Octopus system. 

9 Discussion with J. Fletcher, LLL, 5/89. 


NERCOMP system, set up by Dartmouth University as an outgrowth of the Dartmouth 
Time Sharing System, by the late- 1960s Unked a number of smaller academic institutes 
throughout New While relatively slow and unsophisticated, this was periiaps 
the first time-sharing network to be operated on a pay-for-itself basis." 

Roberts came to ARPA in late 1966 and commenced developing plans for 
networking to link computers. R. Taylor, head of IPTO at that time, had a background and 
ideas similar to Licklider's about the benefits from developing man-computer interactions 
on a broad front He was anxious to involve the 15-20 computer researchers supported by 
ARPA in planning the initial ARPA network, soon to be called ARPANET. An informal 
working group made up of most of these researchers helped assess and plan different 
possibilities for communication links between their research computers, which were of 
many different types and used generally different operating systems and communications 
control programs. 

This group soon concluded that a distributed, multinode network was needed, 
which could be linked by leased telephone lines with faster switching and wider bandwidth 
dian the common cairier switched voice network. A key suggestion was made by W. 
Clark that small intermediate computers, between the "host" computers resident at each 
users' location (or node) and the communication lines, could remove some of the burden of 
programming each different host computer to interface with the conmiunication lines.^^ 
Communications in the ARPA network was then envisaged as taking place among tiiese 
small computers, later called "interface message processors," or IMPS, in a distributed 
communications network, and between IMPS and host computers. A "hot potato" routing 
scheme, discussed by Baran (about whose work Roberts apparentiy was now aware), for 
handling message segments or "packets" was adopted initiaUy for the new ARPA netwoik. 

10 R. Hargraves, Jr. and T. Hum, "The Dartmouth Time Sharing Network," Chapter 11 in Computer - 
Communication Networks, N. Abrahamson and F. Kuo, eds. Prentice-Hall, 1973. 

11 "In at the Beginnings" by P.M. Morse, MIT 1977. p. 355. ARPA apparenUy provided some 
assistance to Dartmouth for this system, AO. 1075 of 8A57. 

12 Sec "Expanding AI Research and Founding ARPANET." by L. Roberts, in Expert Systems on 
Artificial Intelligence. T. Bartec. ed.. Sams, 1988. Roberts mentions that McCarthy and Minsky of 
MTTs AI group iniaally opposed the idea of others sharing their computer resources. 

13 Toob for Thought, by H. Rheingold, Simon & Schuster. 1985, p. 216. A similar suggestion had also 
been made by Davies. 


IMP routing schemes and algorithms were changed and improved several dmes in the 
ARPANET project, becoming progressively more complex and "intelligent "i"^ 

Robens and his co-workers outlined their rather detailed plans for ARPANET at a 
computer conference late in 1967, A very similar UK NPL plan was presented at the same 
conference, but based on a higher (1.5 Mbit/sec) communication line speed. Discussions at 
the conference influenced ARPANET to use 56 kbit/sec line speed for the "backbone" 
system, a higher transmission line speed than previously planned.15 The objectives of the 
ARPA program stated at this meeting were to develop and test computer-communication 
techniques, and to obtain benefits and economies of resources sharing for as many as 
possible of the then 30-odd ARPA contractors in the IPTO program.i6 It was envisioned 
that short data sets of the type generated in terminal-computer interactions would have to be 
handled by the combined computer and transmission line network with an overall 
transaction time less than the desired human interaction time of about one second. Very 
low error rates were also desired because of the high accuracy required for data 
transmissions between computers, and for this purpose an error-checking code was added 
to each packeL^^ Further network bandwidth requirements came from the desire to have 
remote interactive graphics capabiUty. For this purpose, desired end-to-end bandwidtiis 
had to exceed 20 kilobits/sec. The initial number of users was selected as 15, large enough 
to involve many researchers to help design data formats or protocols together with the 
operating procedures for the network, have interactions between many different kinds of 
cotnputers, and have enough traffic to be able to make meaningful statistical measurements 
and analysis. 

2 . Early Development and Experimentation 

A detailed specification along the lines presented by Roberts in 1967 was set forth 
in an ARPA RFP in 1968, Many major computer manufacturers chose not to bid, 
apparendy because they did not then make minicomputers of the type required for IMPs.»8 

1 ^ Computer Networks, by Andrew S. Tancnbaum, Prentice Hall 1988, p. 289. 

15 "The Evolution of Packet-Switching.- by L. Roberts, Proc. IEEE Wol 66 1978 P- 1308^ 
A^^r speed is a fraction of Se line speed, depending on characteristics of messages and 

16 Roberts later estimated that the savings to the IPTO program was a factor three over Nj^at wouW have 
been required had each contractor been suppUed equivalent computers of their own. Roberts, 

17 -The ARPA Network," by Lawrence G. Roberts and Barry D. Wessicr Ch. 13 in Computer- 
Communication Networks. N. Abramson and F. Kuo. eds.. Prenuce-HalU 1973. p. 4W. 

18 L. Roberts, ibid., 1985. 


Bolt, Beranek and Newman (BB&N) won the contract to design the software for the 
"interface message processors" (IMPs).!^ The IMFs were initially based on a modified 
HoneyweU 516 computer, later, more capable IMPs used BBN designed computers. The 
first few IMPS were built and installed within a year.20 DECCO, a contracting unit of 
DCA in communication services, was given initial responsibility for leasing 56 kbit/sec 
lines, because of favorable government rates. Progress was facilitated by AT&T setting up 
a special unit for dealing with problems of interfacing with the ARPA network for this 
purpose.2i ARPA also contracted with the Network Analysis Corporation (NAC) for 
assistance in designing the "topology" of the network.22 

A "Network Working Group" of key contractors and ARPA managers was set up 
to help design the initial system, especially the software "protocols" needed for 
standardized forms of communication among IMPS, between an IMP and a host, and 
between hosts. In less than a year BB&N had a 4-"node" initial ARPA network, soon 
named ARPANET, set up and running. While inter-lMP conununications were going 
well, the intercomputer links took longer to achieve satisfactory operation. A very 
important feature was that ARPANET was operated from the beginning as a scientific 
experiment, making measurements of important quantitative features and publishing 
results.23 For this purpose one of the key nodes from the beginning was at UCLA under 
L. Kleinrock, with the rcsponsibiUty of gathering data and making analyses. Soon after 
ARPANET started, a "network control" was set up whereby BB&N could remotely 
monitor performance of any IMP and identify and "fix" software problems. This remote 
control of software proved important for econonoic and efficient network operations, and 
for other applications. 

In 1969, a number of other private computer communication systems began to be 
operated, including the SITA system for international airUne reservations, which used 

1 9 A.0. 1260 of 6/68 for "Interface Message Processors." 

20 -History of ARPANET . the First 10 Years." BB&N. p. 24. Software for the "^^was at first 
regarded as proprietary by BB&N. but DARPA ruled that this had ? oftl^^"! ^'i^,^^^ 
S^Xomputers in the PubUc Interest: The Promise and Reality of ARPANET." By D.S. Bushncll 
and Victoria B. Elder, George Mason University, Fairfax. VA. 1987. 

21 BB&N. ibid. 

22 AO # 1380 of 1/69 for "Computer Networic Modelling and Measuremems." 

23 Appaiendy. the French Cycladcs packet-switching system, in operation a bit later, also published much 
of its perfonnance data and associated analysis. 


packet-switching together with voice, and TYMNET for TYMSHARE. one of the large 
time-sharing service companies. These netwoiks involved routing and switching principles 
somewhat different from those used in ARPANET.^* Retrospectively, Roberts points out 
that all these developments were probably due to the fact that P969 was the year when the 
cost of computing fell below the cost of communications for computer-communications.25 

The distributed ARPANET that evolved attempted to achieve the general objectives 
of minimizing costs and maximizing the probability of successful and adequate message 
transmission. In this early growth phase problems of designing such a network began to 
be recognized. One important issue was the optimizing of network topology for these 
objectives.26 The topology problem was not fully solved, but cvenmally approached by 
successive adjustments to an approximate solution. Other problems were routing and flow 
control, taking into account the levels of traffic, capacities of links, and cost Kleinrock 
states that while a number of these problems were and are still unsolved, the network 
operates quite successfully due to the high degree of adaptability of the system and its 
operators. ^7 

Use of the IMPs allowed a degree of standardization of message formats or 
"protocols" over the long communications lines, while reducing the software requirements 
on the host computer operating systems. It was soon found that IMPs should be designed 
to support several hosts in a time-sharing node. Host to host communications via the IMPs 
proved more difficult than expected, and further "interfacing" between host computers and 
the netwoik dirough additional small computers proved necessary in some cases. 

In addition, a need arose among groups without computers of their own to access 
computers through terminals. In 1971, responding to this need, a "Terminal Interface 
Processor," or TIP was designed which allowed direct access to IMPs and so to the entire 

SITA was characterized by BB&N, ibid., as surprisingly sophisticated for its time but not well known 
to the DARPA computer community. See also "TiTMNET I: An Alternative to Packet-Switching 
Technology." by J. Rindc, p. 594 in Satellites, Packets and Distributed Telecommmications. Roy D. 
Rosner, Ed. Lifetime Learning Publication 1981, p. 594. 

25 L. Roberts, Proc. IEEE, ibid.. 1307. This is the cost given the previous investment in the 
communications lines and line-related facilitites used and based (m the current "tariffs" set by the FCC. 

2fi "Principles and Results in Packet Communications," by L. Kleinrock, Proc. IEEE, VoL 66, Nov. 

27 Ibid. Recently, more "intelligent" IMPS can control routing to more closely approximate diese 


network. Costs of IMP'S in the early 1980's were around $50K and TIPs, which 
gradually also absorbed IMP functions, about $100K .28 

3. Demonstration, Transfer, and Initial Applications 

By 1972, having gained considerable experience with ARPANET, ARPA decided 
to stage a pubUc demonstration of its capabiUties. It took nearly a year and considerable 
shakedown effort to arrange for this, but at the Washington International Computer 
Conference in November, 1972, the demonstration, orchestrated by R. Kahn (then of 
BB&N), was very successful. This demonstration Unkcd, via ARPANET, some 25 
terminals at the conference location with a variety of computer resources. In 1973 
ARPANET was made available to DoD and its contractors, who became a fast-growing 

After this successful demonstration of the ARPANET technology, an approach was 
made by ARPA to AT&T to take over operation of ARPANET as a pubUc network, with a 
view that such a "utiUty" could serve commercial, research and military users. However. 
AT&T, which also was opening circuit switched services for data transmission at the time, 
declined.29 similar discussions were held with other common carriers, but a GAO report 
raised the issue whether ARPANET, a government-funded system, should not be first 
offered to government agencies,30 After the GAO report, ARPA commissioned wide- 
ranging studies of the utiHty of ARPANET which laid the basis for high level discussions 
in DoD. leading evenmally to negotiations with DCAJ^ 

The mission of DCA was to provide communications for the military and it was at 
first reluctant to operate a research network such as ARPANET which also involved non- 
miUtary users, and which had at the time no provisions for security. However, within 
DCA no one in authority voiced major objections to taking over responsibUity for 
ARPANET.32 There were, also, several other factors affecting DCA's actions regarding 

2S What Can be Automated?, MIT Press, 1980. p. 383. 

29 In 1976 AT&T used packet switching extensively in its COS between its switching nodes, to control 
communications, and later also offered a form of packet switching service to customers. See e g 
"Evolution of the IntcUigent Telecommunicauons Network." by John S. Mayo, Science, vol. ^n, 
1982. p. 831. A display of telecommunications "teeakthroughs" in this article, however, does not 
include packet-switching. 

30 Discussion with R. Kahn and V, Cerf. 5/89. In fact. ARPANET technology had been picked up 
quickly by NSA. 

3 i P. Baran, who had done the earliest studies of packet switching, participated in these studies. 
32 Discussion with E. Hoverston, 5/89. 


ARPANET: (1) there was a growing number of military nodes of ARPANET; (2) ARPA, 
in order to be able to share classified data over the network undertook to develop, with 
NS A, a "private Une interface" (PLI) device allowing end-to-end ARPANET encrypaon;33 
and (3) internal studies by DCA of the next generation defense data communication system 
indicated the desirability of using packet-switching technology. An agreement that DCA 
would take over operating responsibilities of ARPANET was effective in mid 1975, and 
allowed DARPA to continue its research programs on the network as a 'DoD sponsor." 

ARPANET grew rapidly in number of "nodes." and in traffic volume in die first 
few years. Figures 1 and 2 show the ARPANET network at early (1970) and later (1985) 
stages. Early estimates had been that the traffic growth would be exponential and that 
network capacity would soon be saturated. It soon turned out that the growth flattened out 
and that the host computers were saturated before the network.^* In the mid 1980's. 
however, network congestion was common.35 Also, early estimates were that message 
length distribution would be bimodal, with many short messages and a smaller number of 
large messages.^^ Eventually, short "electronic mail" messages dominated. 

BB&N, with the ARPANET experience under its belt, was encouraged by DARPA 
to set up a public packet-switched data network under the new FCC rules.37 bB&N set up 
a subsidiary, TELENET, to do so, and Roberts left DARPA and joined HELENET soon 
afterwards. Apparently, however, it took neariy two years to raise enough venture capital 
and to get FCC approval to launch die new network. TELENET started operation in 
1975.38 In a few years TELENET grew to serve about 200 nodes in different cities. 
TELENET incorporated "Virtual Circuits" and ARPANET "datagram" technology.^' 

33 AO 2755 "Net Encryption" of lli74 and A.0. 3092 of 8/75. 

34 BB&N, ibid., p. in-72. This was apparcnUy due to a rapid adaptation by the users. BB&N. ibid., p. 
m-74. ' 

35 Toward a National Research Network^t^axionalAcddtmyoiScitncts, 1988,p. 11. 

36 Kleinrock, ibid., p. 1320. 

3'' D. BushneU and V. Elder, ibid. 

38 "Electronic Post for Switching Data," New Scientist, 15 May 1976, p. 351. and "Three Decades of 
Contributions in Science and Technology," BB&N, 1988. p. 10. 

39 Virtual circuit technology with flow control apparently was pioneered by the French RCP packet- 
switching system. See Roberts, Proa IEEE, ibid., p. 1309. 






0E-64r) I TSP ) 




Figure 1. An llluslrailon ol the proliferation of networks "^"^^ 
worldwide. (Modified from IEEE Spectrum. Vol. 25, No. 2. February 1988. 

pg. 56.) 


Figure 2. Evolution of the ARPA Network, (a) December 1969, (b) December 
1970, (c) September 1971, (d) August 1972, (e) November 1974 (from Howard 
Frank, "ARPA Network," Proc, IEEE, Vol. 66, No. 11, November 1978) 


Figure 3 shows the worldwide proliferation of network activity from 1972 to 1975. 
This can be credited to several factors: (1) the impact of the economics of 
computing and of communication, worldwide: (2) in the U.S.. the FCC decision to permit 
value-added cairiers to compete with Ae established canieis; (3) that the technology did not 
require any major technological breakdirou^; and. perhaps most importandy. (4) the 
impact of the existing operating ARPANET and the published scientific information about 


4. Expanded Defense Application 

From the early 1970s into this decade ARPANET packet switching technology has 
been the basis for the development of defense-wide systems for data communications. 
While several appUcation efforts started in the early 1970s, the development of this 
defense-wide capability began with the miUtary nodes of ARPANET which were already 
heavy users of ARPANET through the 1970s. Starting in 1971 interactive networking 
efforts in both the command and control (WIN ) and inteUigence (COINS) arena began as 
experimental extensions of ARPANET packet switching technology. In botii of these 
efforts individuals who had been directly involved in the development and use of 
ARPANET carried these concepts into their specific highly classified user environments. 
Through the 1970s, these experimental prototype netwoiks grew into and were accepted as 
operational systems within the confines of the security limitations of these classified arenas. 
Attempts were made starting in 1972 to introduce some packet switching into a planned 
replacement of the AUTODIN system for DoD message and data communications. This 
effort. AUTODIN U, was judged to be unsuccessful, and in 1982 a decision was made to 
implement an alternative approach for interactive data communications, die Defense Data 
Networic (DDN) based expUcidy on ARPANET incorporating the MILNET and die WIN 
networks. These developments, described below in more detail, proceeded in parallel, but 
not in isolation. There was early recognition of the desirabiUty of interUnking die 
independent network developments, but also an appreciation of the difficulties of doing so 
given the differing levels of security diis would cntaiL While considerable progress has 
been made, the intemetring of the DoD ARPANET-based packet switching netwoiks still 

is not complete. 

The transfer of operational responsibility to DCA in 1 975 highlighted a dichotomy 
in die character of ARPANET as a dual purpose system-both a research network and an 


Tlma iinM of notable computor nttwortcs 

rtn iwi 9n iwb ht? wrt iwi 
I I 

i 1— t H 

H 1 I I 1 1 I I t i ll' 

ACONir. AkMamnotM Comeuttr Mwa 

ASCNR Au»«»H«n CompuW 8ci«nn NalMili 
AUUANR AuxroMttn NatwotK 

auNrr u ib a*nw\ Natwon 

trran (tl T(m« NvTwofk 

COSAC: Comfnuniotiona un* ConnaCttOflf 

CSMMR Commommitn SCMnilfte fc liMuMft- 

tJ AMMicn OfgantntMn Nvtwortc 

OUR Computar f*"*^ 

DON; 0«fm— P«tt W»iww * 

DCCMR Bhmc dec CAffinMnng NaiwoM 

DSlKiM OOMmiNnl iwttMTK M N«v 2Maii4 
■ARM: euiDpaan MaMmtc nimwH Nataerk 

tut. EufopMA irrtomwito NMWMk 

EUmC EwiopMn UMx MMwora 

HAIINR Hooo Kono » nm ii T h 

HiAMR M»c»w bhieittton Authortty iimmqM 
HtMR Hlon Emtot WO W CI IIMMfK 
HMIMCT i. tt: HaMHMtfMr MNtttM 
t*M VMR VkttMl Natwark 
mwurt Mttuto t U i>Bwa<i n«ea M yo t tiw 
macMk tnMWMiwnai O iBnaaan le r aandwd 
tzkiUn-Opan tyMnw iMMeMMtet 

PMHOUMR »*or«e UoHwiWy ^ 
NtAwc N«uanaiSBiMMFaunMtMnN«Mik 

PACHR FteHie N«t«ark 

RMOKoM; lUnguiwi KompuMr MMyM 

RnmUfc fOMUB «M UntMmtlis at m la te- 

SON: »yatam Otiaiop w an t HafwoM 
tlMOMt: Sotanoa btoMiMrMO RMaaRA C«w 

MMR ScMumDaniar IntofmaUon NaNMk 
IWMtSoaca WtyatcaAnaiyaiaWawwnt 

MR loutn PMine Eduntton ana ll»> 


SUIIVMR 0«W» vHMnttr nawork 
UMA: UiiMrattlil»M«a Auatrta 
UMWint Nortic wnMntty namoffe 
waiNR Uaara- NaMPfk 
UUCP; UMi ta iMi Cow 
Xam cut: CoivoraH mtamrt 

M nitaaHi Ataoelii pM M Mem 


Thepaa/ht jmm hanMm tim 
Awnter o/ Rcrwohki soar dlro- 

Jio/otx d€V€ioped by one or 
more older networks; these are 
indieaud by the verticai dashed 
lines eonn*etint networks. 
(More TteenUy. some networks 
have begun to use other pro- 
tocois, partkulariy ISO-OSI tntd 
Arpanet standards.) Solid verti- 
cat lines between networks indi- 
tatesystems under dosdy reiat- 
ed administrations. Dashed 
horijontal tines indicate pro- 
tocois or demonstration rysiuns, 
rather than operational not- 
works. Networks in itatk s are 
internets — seveml networks tied 
together that use the soma maa- 
mission protocols. 

Figure 3. An Illustration of the Proliferation of Networks Used by Researchers 
Worldwide. (Modified from IEEE Spectrum. Vol. 25, No. 2, February 1988, 

pg. 56) 


unclassified defense network for military users. With increasing use by military users for 
•operational," as opposed to research applications, this dichotomy raised organizational 
concerns within DCA."*® 

...ARPANET has had a dual character. On the one hand, it has existed as 
an operational network serving a wide variety of users. On the other hand, 
it has served as an experimental testbed for research on packet switching. 
...ARPANET is operational DoD facility, used solely for government- 
related business. The operational users require reliable, consistent network 
service ... and... attention paid to security and privacy. 

With tiie creation of DDN in 1982, tiiesc military nodes were split off from ARPANET as 


The Worldwide Military Command and Control System (WWMCCS), under the 
auspices of the Joint Technical Support Agency, purchased an ARPANET-type system 
from BB&N for tiie Prototype WWMCCS Intercomputer Network (PWIN). This was "an 
experimental program from 1971 to 1977 to detennine tiie operational benefit of 
networking and to identify the characteristics needed to support military operations."^! 
WWMCCS, whose communications were being provided by DCA, had been procuring 
H6000 series computers for DoD's major Unified and Specified Command Centers, This 
provided the equipment compatibility for the development of intercomputer 
communications within WWMCCS, a capability that was seen as essential. 

The tests of PWIN proved sufficiently successful, despite some problems, that it 
became the basis for the much larger "WIN" system. Six initial WIN sites in 1977 
increased to 20 sites by 1981. However, problems in the technical and procedural aspects 
of systems performance led, in 1980, to a major program to upgrade hardware, software 
and reliability.^2 This upgrade was completed in 1983.^3 As will be discussed below, in 
1982 tiie DDN, initially called tiie "WIN/ARPANET replica," was built upon tins base.^ 

*0 T. Harris, et al., "Development of the MILNET," CH1828-3/82, IEEE. 1982. p. 78. 
*1 Modernization of the WWMCCS Information System (WIS), Assistant Secretafy of Defense, 
(Conununicattons. Command, Control, and Intelligence), 19 January 1981, p.7. 

Ibid., p.7 and p. 39. 

Defense Science Board, Defense Data Network, Office of the Under Secretary of Defense for Research 
and Engineering, 1985, p. 3. 

Hearings before Defense Subcommittee of Committee on Appropriations, HOR. 96th Congress, 1st 
Session. Part 6, 1979, p. 253. 



In 1965, the National Security Agency (NSA) began the Community On-line 
Intelligence System (COINS), an "experiment in exchange of intelligence information 
throughout the intelligence community." COINS was initially a store-and-forward network 
which became operational in 1973.^^ From 1973 to 1977 COINS was upgraded from a 
storc-and-forward to a packet switched system based on ARPANET technology. The 
packet switched network, COINS H. was declared operational in 1977.'** The following 
were seen as the features and advantages of the new ARPANET-based COINS:*^ 

• The star network switch has been replaced by a distributed, packet-switched 
communications system modelled after ARPANET. There is no longer a 
single point of failure. 

Hie protocol set has been enlarged to include interactive operation. 

• Host systems are attached to the network via front-end processors, which 
execute the network protocols. The hosts are thus freed from a substantial 
(and increasing) network overhead burden. 

• The network can be accessed from terminal concentrators which are not 
directly associated witii any network host. Given proper authorization and a 
secure environment, any terminal can access COINS from any location. 

The COINS initial store-and-forward configuration was established at die Defense 
InteUigence Agency's (DIA) Arlington Hall faciUty and linked to NSA. In 1973, through 
1977, additional intelligence conmnunity hosts were added to the packet-switched system 
and in 1978 the first terminal concentrator permitting access to the network from points not 
associated witii a host computer became operational.^^ By 1980, while die system was 
generally operational, it was constrained by accessibility problems due to the age of some 
of the computers, lack of necessary interactive protocols between some of the network 
components, and die mixture of non-standard front-end processors. A key limitation was 
die lack of a multi-level security capability, restricting access to the SI/TK level. "Most of 

*5 com Long Range Plan, Part U COINS Network Architecture for the Long Range Plan, COINS 
Project Management Office. NSA. Ft Meade. Maryland. 23 March 1981. pp. 1-2. 

46 Ibid. 

47 Ibid., p. 5. 

48 Ibid. 


the potential inteUigence community users [were] thus excluded from COINS."^^ 
Although the access problems due to both technology and security limitations were still 
needing resolution, it was envisioned that COINS would interconnect via "gateways" to 
several other networks either in existence or then in the planning stage: ARPANET, 
PLATFORM, IDHSC. AUTODIN H. and IAIPS.50 importantly, these intcrconnectivity 
plans were being made under the assumption that the new DoD-wide data communications 
system then under development, AUTODIN H, would become operational. The failure of 
that development and the difficulty of achieving acceptable multi-level security gateway 
links between COINS and other DoD inteUigence networks have delayed the envisioned 
inter*netwock connections. 


In 1972 the first plans for the new DoD AUTODIN H lelecoimnunications system 
began to be laid^i This was parUy in response to requests originating from the new 
Assistant Secretary of Defense for Teleconomunications, Dr. Rechtin (the ARPA director 
during the early phases of ARPANET), who had "tasked the Director. Defense 
Communications Agency (DCA) to make reconomendations concerning the provision of a 
family of Defense Communication System (DCS) switched services to fulfill computer 
communications requirements for the DoD."» In addition the Joint Chiefs of Staff, in July 
1972, tasked the Director. DCA to prepare a plan to satisfy WWMCCS ADP 
communications requirements. DCA smdies of users' requirements were then ongoing for 
a new system to replace AUTODIN I. Essentially a teletype message switching system 
with store-and-forward capabiUties, AUTODIN I was recognized to be slow and unable to 
handle interactive computer traffic, for which there was increasing demand in the DoD. 

The computers at military installations which were to be Unked by DCA were of 
several different types, often with their own software. Large doUar and training economies 
appeared possible if they could be Unked together via a network in which, like ARPANET, 
these computers could communicate with one anotiier and be able to share software and 

^9 Ibid, p. 8-9. 

50 Ibid., p. 11. 

5 1 "The Autodin II Network," by Col. A. Siathopoulos and HJ. Cally, EASCON-77. IEEE, 1977. p. 

52 Ibid, 


other resources." A panel, mcluding some from the ARPANET community, was called in 
for assistance by DCA and recommended IMP-type interfaces and ARPANET-like 
protocols for the network and the "backbone" long haul communications circuits. 

Despite the recommendations of the advisory panel to use ARPANET technology 
and protocols, the AUTODIN system detailed in the System Performance Specification 
showed substantial differences between the characteristics of AUTODIN 11 and 
ARPANET.54 A key difference was that AUTODIN U employed only a very few (initially 
four and planned eight) central nodes into which data would be directed and rerouted, 
requiring very large message storage capabilities in each central node. Moreover, each 
center required many personnel cleared to the SVTK level and TEMPEST secure, guarded 
facilities. This architectural aspect of AUTODIN II substantially reduced the effectiveness 
of the packet switching capabilities of the intemode communications. The recommendation 
of DCA was based on the fact that tiiere was already a large inventory of older AUTODIN I 
equipment, and switching over to an ARPANET based packet switching system was seen 
as a very costly approach, given this installed base. 

Moreover, the technique for assessing the security classification of messages used 
an approach that was cumbersome and manpower intensive, yet DCA was not satisfied that 
its security requirements could be met adequately by packet switching. The individual 
nodes were very large operations, with large data storage systems and had sizeable 
manpower requirements to enforce security since the data within a portion of each node had 
to be in the clear for routing purposes. Multilevel security for AUTODIN n was based on 
a software "security kernel" approach, which proved to be difficult to implement and certify 
as sufficientiy trustworthy for data above the secret level. 

AUTODIN H construction commenced in 1977 and proceeded at a very slow pace, 
even with only 4 nodes in die initial phase. The difficulties encountered in implementing 
this system led to a major review tiiat led to AUTODIN n being superseded by an 
altcmative approach, the DDN:^^ 

As a two year program for initial implementation stretched to four and a 
half, a growing number of problems and uncenainties about AUTODIN II 
were encountered. In July 1980, an OSD review group was established to 

5 3 "History of the ARPANET." BBN, ibid., p. U-A, 

Stathopoulos. ibid., p. 8-lC. 
55 Report cfDrfense Science Board Task Force on AUTODIN II, Office of the Under Secretaiy of Defense 

for Research and Engineering, December 1982. p. 3. 


review the system... (which) ... considered the cost, secunty, performance, 
and survivabiUty of AUTODIN n..,.[T]he group also explored available 
options if AUTODIN U failed. Principal among the alternatives considered 
was an expansion of the WWMCCS Information Network and ARPANET 


There were growing concerns about and criticism of AUTODIN H because of the 
generally slow pace of progress, the lack of potential to meet growing needs, and most 
importantly, costs.56 SurvivabiHty of the system, which was estimated to be low for the 
AUTODIN n nodes, was also a concern. Because of the necessity for a digital DoD 
network to provide interactive service, the Assistant Secretary of Defense for C^I 
(ASDC^D tasked the Instimte for Defense Analyses (IDA) to develop an alternate (or "back 
up") design in case the AUTODIN U system problems proved insurmountable.^^ 

The design produced by IDA had two separate networks, (1) an unclassified 
network (called MILNET) and (2) a classified (C^D network which included service for 
WIN, DoDnS (then IDHS), and SACDIN. The design used ARPANET and its packet- 
switching technology, C30s, the updated IMPs. were used in the switches. TCP/IP and 
X.25 or 1822 were proposed as lower network protocols. A key point in the design was 
the use of private line interface (PLD devices (or their successors, IPLIs and BLACKER) 
to provide end-to-end encryption to separate classified users.^s The coUocarion of WIN 
and DODHS sites and the short runs to switches provided economy and the many switches 
provided survivability. 

The proposed network design was circulated and many potential users stated strong 
preference for this design versus the AUTODIN n design. The ASDC3l then tasked the 
Defense Science Board (DSB) to review the AUTODIN problem and the projk)sed 
solution.^ The DSB Task Force recommended the termination of AUTODIN n and its 

56 Hearings before Defense Subcommittee of Committee on Appropriations. HOR, 97ih Congress. 2nd 
Session, p. 91 ff. 

57 The foUowing is derived from discussions in 8/89 with T. Bartcc of IDA. who developed the DDN 

58 A.0 3173 of 12/75 had provided for development of PLI's. 

5 9 Report of Defense Science Board Task Force on AUTODIN II. Office of the Under Secretary of Defense 
for Research and Engineering. December 1982. 


replacement by the Defense Data Netwoik. TTiis recommendation was enacted by Secretary 
of Defense Carlucci on April 2, 1982.«> 

At the same time, ASDC^I also tasked DCA to determine the optimum design for 
DoD. DCA formed three task forces-(l) a group to update and improve the AUTODIN n 
design and explore fumre possibilities and costs; (2) a group to further develop the details 
of the design proposed by IDA and predict fumre developments and a more detailed cost 
estimate; and, (3) a team to decide between the two designs. 

The result was a choice of the ARPANET technology plus NS A/DARPA security 
features. AUTODIN H was canceUcd and the IPU and BLACKER projects were initiated. 
A DDN office was formed at DCA under Col. Heidncr, who had headed the winning 
design teaxxL 

The planned evolution of the DoD nctworic from the 1982 Defense Science Board 
Report, shown in Figure 4, "consists of the evolution and expansion of existing and newly 
estabhshed networks based on ARPANET technology and their ultimate consoUdation into 
an integrated network suitable for use at multiple levels of security."*^ DDN was planned 
to be a more survivable system with a much larger number of distributed nodes and links. 
The use of ARPANET technology permitted easy expansion of the networic. By this time 
the experience with operating ARPANET and the open scientific data published about it had 
also built confidence in the technology. 

Because the BLACKER and IPLI were in development, the DDN was originally 
designed in separate pieces, including MILNET, ARPANET, WIN, DODHS, "Secret 
Net," etc.62 This was as planned, however. Merging the classified sections has been 
delayed because of BLACKER delays and NSA's decision to continue only BLACKER 
and not the IPLI program. Apparently the problems of achieving adequate multilevel 
security, witiioul the high expense of a large number of IPLFs, has proved more difficult 

60 The cancellation was not sudden but had been planned for some lime. It took place one day after the 
formal contract completion date, to minimize ovcraU costs. Discussion with V.Cerf and R. Kahn 

«1 Final Report Defense Science Board Task Force on D^ense Data Network, Office of the Under 

Secretary of Defense for Research and Engineering. 30 August 1985, p.2. 
62 Testimony of D. Latham. Deputy Assistant Secretary of Defense (Communications, Command. 

Control and Intelligence). Hearings of the Subcommittee on the Department of Defense of Uie 

Committee on Appropriauons, Defense Appropriations for 1984, House of Rcpreseniauves. First 

Session. 98th Congress, May 11, 1983. p. 343. 


than anticipated." At present, plans to merge the classified networks have been 
estabUshed and BLACKER testing has begun on operational networks. WoiTies about 
computer "viruses" make interconnection of the classified and unclassified network 

As "[a] first step in the evolution of the DDN," MILNET was established, 
separating out the operational mUitary nodes ftom the ARPANET.*^ MILNET handles 
unclassified but sensitive operational traffic using commercial grade cryptographic systems, 
and until recently had a link to ARPANET thiougji a physically separate "gateway." 











Source: Final Report, Defense Science Board Task Force on Defense Data Network^ 
Offi^of ITuSo^Secitary of Defense for Reseuch tnd Engineenng. 30 August. 


Figure 4. DDN Network Design 

64 Ibid.p.77. 


MILNET was spUt off ftom the rest of ARPANET initiaUy by the TCP/IP software 
protocol developed by DARPA. and effective in 1984. when this protocol was accepted by 
DCA. Secure gateways also Unked MILNET and its European counterpan. Movements 
Information Network (MINET), to classified DCA networks.^^ The MILNET/MINET 
network has grown to approximately 250 nodes reaching "most DoD faciUties around the 
world, stretching from Turkey in the east around to Guam and Korea in the west." 

5 . Other ARPANET Research 

As the ARPANET demonstration and applications in telecommunications 
networking showed the promise of packet switch technology, DARPA pursued additional 
areas of its possible application. These included "Packet Radio," "Packet Voice," and 
"Packet SateUite." In addition, the ARPANET itself became an imponant contributor to 
successful conduct of other DARPA programs, in particular, the AI research program and 
MOSIS, a program to facilitate integrated circuit fabrication research. 

Packet Radio 

Experiments were conducted in the early 1970's to link computer users by "packet 
radio," beginning with the "ALOHA" system linking educational institutions in the 
Hawaiian Islands." The concept of linking computers by packet switching 
communications using radio broadcast rather than conventional lines appeared to offer 
many advantages, particularly for Army mobile systems in the field. Some packet radio 
demonstrations were later conducted witii tiie Strategic Air Command (SAC). Special 
broad band, countermeasure resistant radios were developed for field test at Fon Bragg, 
but proved expensive. Problems with multipatii transmission and interference were 
investigated. Related R&D has continued jointly witii die Army to date. Problems of 
"collision" of messages from many transnutters, characteristic of the radio packet 
environment, were dealt witii by arrangements such as "slotted Aloha," due to L. Roberts 
of DARPA. Packet contention problems in local area networks have been handled also by 
techniques related to tiiose used in ALOHA.^' 

65 D, Perry, et al.. "The ARPANET and the DARPA Intcmct," Ubrary Hi TECH, VoL 6. No. 2. 1988. 
p. 56. 

66 R. Kahn who joined DARPA in 1973 led this packet radio dcvelopmeni effort. "Adyanoein I^kct 
Radio Technology." by R.E. Kahn, et al., Proc. IEEE. Vol. 66, 1978, p. 1468. also The Aloha 
System," by Abramson, et al., in Computer-Communication Networks, Abramson and Kuo, eos., iDiQ. 

67 "An Introduction to Local Area Networks," by D. Cline et al., Proc. IEEE, Vol. 66, 1978. p. 1497. 


High costs of the packet radios developed for Army field use were addressed by a 
special joint DARPA-Army effon. However, the Army decided recenUy to save time, 
some expense, and its TRI-TAC programs by "jumping" the R&D process, and as a "non- 
development initiative" purchased in 1985, for field trials of "mobile subscriber equipment" 
(MSE), a version of the "RITA" field radio system which had been developed by the 
French in the mid to late 1970's. The U.S. Army version of RITA is apparently a circuit 
switched system, with a central control node.^8 An upgrade to incorporate packet- 
switching is expected in the 1990's.69 Also, the Air Force is installing an electromagnetic 
pulse hardened packet-switched radio system, die groundwavc emergency network 
(GWEN), for missile warning centers, command centers and strategic force bases,™ 

A spinoff of DARPA's efforts in packet radio was made to speed up the solution of 
some logistic problems of the 82nd Airborne Division. Very rapid adjustments of space, 
weight and lift capabilities are faced when loading this division for different missions 
when, as typically occurs, changes have to be made because of aircraft and equipment 
avaUabiUty. The AALPS computer-based system for loading the division was developed 
by SRI witii support from die DARPA packet radio program. With a computer terminal on 
die airfield, a mainframe computer which can run AALPS could be accessed by radio. 
Adjustments could then be made on the airfield, in near real time, according to dynamically 
changing availability of aircraft. After a number of trials including one experiment using a 
group of sergeants making manual calculations as competition, AALPS was adopted by the 
82nd Division and is now part of their regular procedure for rapid deployment 

Packet Voice 

In the early 1970's experiments began using ARPANET packet switching 
(digitized) voice and combined data and voice communications, using both lines and packet 
radios.''^ Packet digitized voice has advantages for encryption and efficiency in military 
communicauons, but loses much of an individual's speaking (and so identification) 

68 Discussion with Col. W. Stevens, IDA. 3/89. RITA apparenUy can have a packet switching wpabUity 
as did its competitor, the UK's Ptarmigan, but this feature is not now being exploited by the Amy 
system. See A. Wohlsietter and R. Brody. "Continuing Control...", Ref. 3., pp. 176-177. 

6^ Jane's Military Communications, 1989, p. 810. 

70 A. Wohlstetter and R. Brody, "Continuing Control as a Requirement for Deterring," ibid., p. 177. 
''I Discussion with V. Cerf, 5/89. 

72 "Experience With Speech Communications in Packet Networks." by Clifford J. Weuistein and Joseph 
W. ¥o[gie JEEE Journal, on selected areas in communications. Vol, SAC-1 No. 6, See 1983, p. yoJ. 


characteristics. Delay times for presently available bandwidth circuits also proved 
troublesome. ApparenUy, satisfactory voice and data communications, with many users, 
will require wider band circuits and faster switches than initially used by ARPANET.''^ 
Work along these lines, over wideband, higher speed links, has intensified recently and has 
involved active participation of the "common carriers," such as AT&T. 

Packet Satellite 

ARPANET wideband satellite packet switching links were set up with Hawaii, 
Norway and London.'^ SatelUte packet switching investigations led to a commercial 
service offered for a whUe by Western Union, but now shut down. SatelUte packet 
communications apparently have found use primarily in applications which are less 
sensitive to transmission delays.^s SIMNET, a graphic simulation system which uses 
satelUte packet switching for training widely separated Army tank crews, has had growing 

Local Area Networks 

"Local area" networks (LANs), with limited geographic distribution and greater 
bandwidths than ARPANET, began in the mid 1960's. One of the earUest was the 
Lawrence Uvermore Laboratory's OCTOPUS system, mentioned above, which was based 
initially on concepts pubUshed by project MAC^^ lLL developed their own dynamic 
switching software (with some limited packet switching capabiUties) to link their several 
different types of large computers dircctiy to each other and to ieniiinals.''8 

In the early 1970's Xerox constructed Ethernet, partly based on ARPA's packet- 
switching technology developed for packet radio.''^ Ethernet soon became a commercial 
success. Local area network systems, based primarily on ARPANET technology, also 
developed rapidly in DoD agencies. The growth of LANs and other networks within DoD 

"^3 L. Roberts, unpublished, 1985. 

74 NORSAR was the teminal in Norway for data transmission to the seismic research center of D ARPA's 

75 Discussion with Dr. V. Ccrf, 5/88. See also "ARPANET Hitches a SateUiie Ride," by S. Blumenthal, 
Communications Systems Worldwide, Sept 1985. 

76 i>iscussion with J. Orlansky. IDA, 3/88. 

77 Discussion with J. Fletcher, LLL. 5/89. 

78 Phcrson, ibid., p. 229. 


brought a need to fonnulaic protocols which had provision for security. DARPA led the 
successful effort to define the TCP/IP protocols for multilevel security. 

ARPANET as a Research Tool: AI and MOSIS 

In providing interactive computer communications among researchers, ARPANET 
contributed to several ARPA computer-based development efforts. One successful effort 
to exploit ARPANET was the intensive use of "electronic mail" and a form of 
teleconferencing to develop the AI language, Conmion USP, Still another successful 
ARPANET exploitation has been made in MOSIS, a system to expedite fabrication of 
integrated circuits. A central faciUty for MOSIS is provided by the University of Southern 
California's Information Sciences Institute. 

As described by Newell and SprouU, MOSIS allows integrated circuit designs to be 
transmitted to a fabrication facility:^ an electronic mail message describing in a text form the geometry of 
the several masks that control integrated-circuit fabrication.... MOSIS uses 
die network to allow a great many designers to share access to fabrication. 
Moreover, the system is able to combine several separate designs onto one 
chip (a so-called multiproject chip) in order to reduce fabrication cost. 
Centralizing fabrication services in this way simplifies interactions wiui 
vendors and frees die chip designer from a great many troublesome details. 
An important advantage is the avoidance of dealing with a human 
bureaucracy (die alternative organization technology for managing the same 
process), which tends to become unresponsive, error prone, and hard to 
control.... [The network] becomes an integral part of a larger computational 
enterprise. The design sent by [electronic] mail to MOSIS is not prepared 
by hand, but is produced by computer-aided design tools for preparing 
mask geometry and for checking the design. 

ARPANET'S Impact on Internetwork Communications 

The value of the DARPA effort to develop protocols for internetwork 
communications was recognized by die international community, and DARPA again played 
a prominent role in the remarkably rapid development of international standards for 
computer-network and network-network commuiucations, such as the CCITT very 

79 R. Taylor, ex-hcad of DARPA s IPTO, went to Xerox and started PARC, where ETHERNET was 
buUL Sec Tools for Thought, by H. Rheingold, Simon & Schuster. 1985, p. 205 ff. 

80 "Computer Networks. Prospects for Scientists." by Allan Howell and Robin F. SprouU, Science, Vol. 
215. Feb. 1982. p. 849. 


similar to the ARPANET TCP/IP protocol. Other related developments, such as "virtual 
links" with individual flow control, originating with the French RCP network, also played 
an important role in setting More recent development in standards have led to 
the International Standards Organization's "Open System Interconnections" protocols, 
gradually being adopted worldwide, which differs from the TCP/IP of ARPANET, but has 
as yet much less working experience. Many, if not most, commercial network systems are 
now based on TCP/IP.82 

Within the research community demand for network capabilities has increased 
markedly, due to developments such as the convenience of "electronic maU," and the desire 
to facUitatc access to supercomputers." The availability of "free" electronic maU on 
ARPANET had a major impact on the style and efficiency of research by its users. Another 
motif comes from the desire for simultaneous processing, e.g., for geophysical research or 
seismic monitoring, of worldwide observations. NSF. in the mid 1980's, set up an 
agreement with DARPA initially to aUow expansion of the number of nodes in ARPANET, 
to include NSF-supponed research groups, and later linking ARPANET to other nets such 
as CSNET.** Network traffic levels apparentiy have increased to the point of frequent 
congestion and less reliable internet performance. 

With increasing demand for remote usage of supercomputers, the need for greater 
bandwidth and higher speed transmission links has led to plans for a new wideband 
network, with corresponding switching speed capabilities. ARPANET, according to 
recent reports, will be replaced by a new 'Ttefense Research Net." with the new range of 
capabilities, also to be run by DCA.85 These new capabUities bring with them also a new 
generation of problems related to the design of the interface processors, switching 
software, network designs, and economics. 

In 1982, L. Robens and L. Kleinrock were awarded Ericsson prizes, die Electrical 
Engineers' version of the Nobel Prize, by the government of Sweden, in recognition of 
their contributions to the technology of packet-switching. 

8 1 Roberts, unpublished, 1985. 

82 V.Cerf.ibid 

o-t V.Cerl.ibia. 

83 Information Technology and The Conduct of Research. National Academy of Science (NAS), 1989. 
Washington, D.C., contains a survey and recommendations for the future. 

84 B Schultz "The Evolution of ARPANET," Dawmarion. Vol. 34, No. 15, 1 Aug. 1988. p. 71, an^ 
Newall and SprouU. ibid., p. 583. Also. Information Technology and The Conduct of Research. NAb. 

ibid., 1989. 
85 Schultz. ibid., p. 74. 



ARPANET was an ARPA initiative, a major result of the "grand scheme" of J.C.R. 
LickUdcr, the first IPTO director, and carried through by his successors, R. Taylor and U 
Roberts. There was software development involved but apparently no technological 
"breakthrough" required for effective implementation of the packet-switching basis for 
ARPANET.8« Roberts describes the impact of ARPANET as "in part a massive and 
evolutionary change in computer technology, and in part a modest and revolutionary 
change in telecommunication technology."" These changes came from the computer 
community and were resisted initially by most of the communications community. 

ARPANET, like the previous time sharing efforts on which it was based, was not 
envisaged as a specifically military development, although it was clearly understood that the 
DoD would be a major user of the technology. This was in accord witii high level 
viewpoints at the time that the U.S. lead in the computer area would be enhanced and its 
national benefit best obtained by a broad R&D effort not tied to specific miUtary projects. 

Perhaps the greatest conmbution of ARPANET was the fact that it was operated as 
an scientific experiment with participation by a highly competent group of contractors, 
whose results and analysis were openly published. This facilitated a broad transfer of 
technology and understanding and provided for estabUshment of confidence in a way that 
would not have occurred if industrial developments had taken the normal course, slower 
and more "hidden" because of inevitable proprietorship, 

Tmiing was a major factor in several respects. In 1972, at the time ARPANET was 
first demonstrated. DCA was in process of smdying the next steps to take with AUTODIN, 
its first attempt at data and message automation. Computer communication was a major 
factor in the smdy. It took from 1972 to 1977 to get AUTODIN XL under contract and by 
the time it reached Initial Operating apabiUty QOQ it had demonstrated many problems of 
cost, schedule, growth potential and vulnerabiHty. It was shut down in 1982. as soon as 
legaUties and economies would allow, and was replaced by DDN. a netwoik based direcdy 
on ARPANET technology. Despite die delays, ARPANET technology probably sped up 
the modernization of DoD communications by several years.88 

86 Roberts, ibid. 

87 Roberts, 1985, ibid. 

88 Discussion with L. Roberts 5/88. 


ARPANET flourished as an unclassified network. When discussions began about 
DCA taking over responsibility for ARPANET, network security became a major issue, 
resulting in a DARPA program leading to die widely used TCP/IP protocol. However, the 
recent experience of the intelligence community and DDN widi multi-level security indicates 
the difficulty of achieving an economic and satisfactorily secure defense network. 

ARPANETs development was well timed technically, economically, politically, 
and in regard to military needs. The economics of packet versus circuit switching keyed to 
the rapid fall in computer hardware costs, and the FCC decisions in the U.S., had great 
effect upon the timing of commercial devclopmcnL These feamres of packet switching 
technology also greatly affected DoD decisions regarding telecommunications. The initial 
commercial success of packet switching has now grown to the billion dollar range. 

The ARPANET evolution was paced, of course, by the external technology 
developments relating to chips and integrated circuits embodied in microprocessors and 
memories. In the same period as the corresponding increase of ARPANET capability, 
there occurred an increase of local computing power at progressively decreasing costs, 
through the development of personal computers and work stations. This development 
effectively reduced one of the major early motifs cited for ARPANET: to make larger 
computer capabilities available more widely and with the economy advantage of doing so 
with a small number of large mainframes. In this sense, ARPANET'S use for more 
efficient use of computer resources does not seem to have been as successful as its use for 
electronic mail. However, this objective has returned to prominence with the advent of 
supercomputers. But to accommodate these computers, the packet-switching technology 
has to be updated to accommodate the greater bandwidths and switching speed required 

The development of local area networks in recent years can be regarded as an 
outgrowth of time sharing and packet-switching. Technology transfer to Ethernet, one of 
the earliest LANs, was facilitated by key people moving from the DARPA environment and 
DARPA supported projects such as MAC to Xerox. 

"Packet Radio" has been picked up commercially to a limited extent and has an 
enthusiastic foUowing in amateur radio. While DARPA R&D on field packet radio has 
continued, the Army decided to buy initially a circuit-switching MSE system based on tiie 
French RITA system for its near-future battlefield communications. Apparently, the 
Army's reasons were mainly economical and political. A packet-switching capability for 
the Army MSE System is expected to be available in the I990's. 


"Packet Satellites," except for "batch" type communication or Umitcd categories not 
bothered by the transmission delay, have not been widely used so far. However, the less 
time-sensitive remote-interactive requirements of computer-aided Army simulation training 
systems, such as SIMNET, can accept the satellite transmission delay. SIMNET is now 
beginning to take hold for training exercises involving Army groups at geographically 
distributed groups throughout the world. 

Very effective and efficient transfer of ARPANET technology took place by 
relocation of key people and involvement of key contractors. As mentioned above, strong 
early impetus toward DoD use of ARPANET technology for its data communication came 
from the new DoD Assistant Secretary for Telecommunications, E. Rechtin, who had been 
ARPA director in the ARPANET gestation period L. Roberts, who got ARPANET going, 
went to BB&N to head TELENET. R. Taylor, from DARPA, went to PARC and got 
Etiiemet going. And BB&N, the key ARPANET contractor, became involved witii, first, 
the WWMCCS "PWIN" experimental system, and later witii setting up DDN. 

The greatest impact of the ARPANET program has been its broad, indirect impaa 
on the greater efficiency of R&D, industrial, and miUtary processes requiring computer 
communications. Initially "free" to ARPANET users, this service is now more subject to 
economic incentives in the various networks. Some of the non-military areas which have 
intensively used packet switching technology include medical research and psychology. It 
is remarkable tiiat tiie facilitation of psychological research was the motif that spurred 
Licklider toward the earUest ARPA effons in time-sharing and ARPANET. 

ARPA outlays for ARPANET, from project records, were about $25M to 1975, 
when the transfer to DCA took place. Including radio and sateUite packet switching, and 
network-related research, total outiays are about $150M to date.^ 

The commeicial packet-switching market is cunentiy estimated at about $1/2B .^o 
DCA's first expense for packet-switching for their WIN/ARPANET replica was estimated, 
in 1983. at about $430M.9i The GWEN packet switching network costs to date are 
estimated as about $1/2B.^ 

8 9 About $40M of this went for packet and satellite radio R&D. 

90 Discussion with L. Roberts, 1 1/89. 

91 DoD Appropriations Hearing for 1984. HASC, 98Ui Congress, first session, part 5. p. 

92 HASC Auihorizalion Hearings, FY 1986, Part 2, pp. 127 and 137. 


















The growth of Artificial Intelligence (AI) in the U.S. can be credited greaUy to 
ARPA support, which buUt upon earUcr efforts by the Services and Academia. ARPA 
support of the development of computer time-sharing in project MAC in the early 1960's 
was largely motivated by the need to develop the computer tools essential for AI, Through 
the mid 1970's. building on this base, DARPA* was the primary supporter of AI research. 
DARPA also promoted large focussed AI application efforts, such as automatic speech 
recognition and image understanding. A number of AI appUcations began to appear in the 
late 1970's, inchiding some for tnilitary systems, largely based on technology and 
technologists supported by DARPA. In 1983. AI technology was incorporated as a key 
con^wnent of DARPA's Strategic Computing Program. 


The name "Artificial Intelligence" was given by John McCarthy to describe the main 
topic of tiie first U.S. meeting in the area, supported by die Services and National Science 
Foundation (NSF) in die mid 1950's.i However, a key paper at tiiat meeting, describing a 
successful heuristic computer-based "theorem prover" given by Herbert Simon of Carnegie 
Technical Instimte (now Carnegie-Mellon University), did not use the term "artificial 
intelligence." AI is usually defined as die technology of making computers do tilings tfiat 
would be regarded as inteUigent There is a great deal of overlap with sophisticated 
automation, witii die distinction being that automation pertains to doing things tiiat are more 

The Advanced Research Projects Agency (ARPA) became the Defense Advanced Research Projects 
Agency (DARPA) in 1972, 

Discussions about intelligent computers go back to the times of Gottfried Leibniz and Udy Ada 
Lovelace. In the 1930's and 1940's Turing's work, and later von Neumann s led to further uitcrest m 
"intelligent" behavior of computers. 


or less routine.2 Thus some types of mines long used by the miUtary had actiyation 
systems sometimes described as "intelligent." 

One of the first large efforts of this kind in the late 1950's was undertaken a by the 
Air Force in the related area of automatic language translation. However, such translation 
was found to be quite difficult and a National Academy of Sciences committee reviewing 
the problem discouraged further efforts,^ In this same time period, there were also sptne^ 
related developments by industry in automated design of engines, and in the business area 
for investment choices.^ 

Some research was supported by the Services in the early 1950s on approaches to 
intclUgent sensors and systems based on the study of neurophysical^ processes, and of the 
operations of the brain. One of the resulting devices, jhe "Perception." was capable of 
emulating some of these processes but to a very limited degree because of the limitations of 
technology. But the growing availability of computers at the time, offered another avenue to 
AI, based more on the logical capabilities of computers, which were not then designed with 
brain-like structures to augment human capabilities. It was this latter approach that was 
followed by Simon, McCarthy and others in the major development of AL 

Mathematical logic was one of the first areas in which researchers turned to 
computers to augment human capabilities. In the late 1950si H. Wong of Harvard was 
able to prove several hundred of the propositions in mathematical logic in Whitehead and 
Russell's Principia Mathemaiica, using only machine programming, without having the 
types of heuristic approaches or structured reasoning tools now associated with AI. The 
limitations and cumbersome nature of such an approach for solving deductive logic 
problems with a computer led to efforts to develop a computer language for processing lists 
of symbols. <r 

Around this same time. McCarthy, then at the Massachusetts Institute of 
Technology (MTT), was grappling with the problem 

...could you have a program that would solve a variety of problems, and 
furthermore take advice in order to improve its performance? So he 
proposed some ideas for a program called die Advice Taker.* a program that 
would have common sense - that is, it would deduce from what it was told. 

2 ArHficial Intelligence, by H. Simon. Davis Lecture. Naval War College. National Academy of Sciences 
publication. 1985. 

3 Ibid. 

4 Ibid. ■ 


and what it already knew, the immediate consequences of any actions it 
might take. ^ 

In order to pursue this problem. McCarthy began working on the progranuning 
language LISP, which built upon and made more general the concepts of the list-processing 
languages of Newell, Shaw, and Simon.^ US? since has been developed into a basic tool 
for AI. While McCarthy's earliest woik on US? was not supponed by ARPA, much of its 
later development and implementations were. 

Beginning in the mid 1960's, ARPA began to support the development of AI. The 
initial ARPA support was indirect: Project MAC at MTT to develop computer time-sharing 
at MTT had as one of its main motifs interactive program writing and debugging needed for 
rapid development of AI.'' The development of MACSYMA, a system to aid 
mathematicians with symbolic computation, by Joel Moses of the MTT AI group, was 
much expanded under project MAC.^ Now a commercial product for a range of 
mathematical symbolic processes, MACSYMA derived, in turn, partly from a symbolic 
mathematics effort at the MITRE Corporation supported by the Air Force.^ 

5 P, McCorduck, Machines Who Think, W.H. Freeman, 1979, p. 215-216 

6 Ibid., cf. A NeweU, J. Shaw, and H. Simon. "Empirical Expiorauon of the Logic Theory Machine: A 
Case Study in Heuristics," Proc. 1957 Western Computer Cot^erence, 1957. 

This emphasis was largely due to the insight of McCarthy who perceived the great importance of time- 
sharing for AI development J, McCarthy memo to P. Morse, quoted in A Century <^ Electrical 
Engineering and Computer Science at MIT. by K, Wildes, MIT Press, 1985, p. 243. See also 
McCorduck, ibid., p. 217. who quotes McCarthy that his first funding for ume-shanng was a grant 
from the National Science Foundation. One involved participant observes, "Timc-shanng is not 
Artificial Intelligence, but Artificial Intelligence demanded ii". P. Winston, The AI Business. MIT 
Press 1985, p. 5. 

8 P Winston, ibid.. "Project MAC-25th Arniiversary." MTT, Laboratory for Computer Sciences 1988 
foldout; MACSYMA was an early chaUenge to the "generalisi" concept for AI devdopment embodied 
in Newell's General Problem Solver (GPS), and was considered by some of MITs AI leadmg 
theoreticians at the time not to be AI. Tlie argument was over MACSYMA's reliance on expert, 
specific knowledge, see P. McCorduck, ibid. p. 229. 

9 Discussion with £. Lafferty. 5/89 . 


In the mid 1960's ARPA became a key supporter of AI in the U.S.W Support was 
given by ARPA to the Heuristic Programming Project of Stanford's Edwaid Feigenbaum, 
a fonner student of Simon s at Camegie-MeUon University (CMU). As opposed to the 
broad, general "laws of thinking" that underlay initial AI conceptuaUrations of Newell's 
General Problem Solver, or McCarthy's Advice Taker concept, the approach of 
Feigenbaum was to develop "expen systems" focussing on real, not "toy" problems and 
designed to capmie and utilize expertise in a narrow domain." 

The "real problem" that was the initial focus of Feigenbaum's work was the 
analysis of the structure of organic molecules. Later caUed DENDRAL. this project was 
supported, in the late 1960's and early 1970's, by ARPA. A concern of ARPA was that 
the project was heavUy oriented toward chemistry and that this aspect should be supported 
by others.i2 The National Institutes of Health (NIH) and the National Aeronautics and 
Space Administration (NASA) became funders of the research for automatic interpretation 
of mass spectrograms and nuclear magnetic resonance spectra to identify chemical 
compounds." After NASA support in the AI area dwindled, DENDRAL was supported 
primarily by NIH. and became a widely used laboratory and commercial product in the late 
1970's. DENDRAL is widely considered to have been the first major successful AI expert 
system appUcation. Development of DENDRAL took place over many years and involved 
extensive cooperation of AI researchers and investigators speciaUzing in other fields." 

AI was first expUciUy called out in 1968 or 1969 as a separate research area in the 
ARPA IPTO research budget" ARPA support was given both to fundamental areas, such 

10 In the early 1960's there were a number of studies and meetings on AI inthe UK. Largely due lo this 
^ *Uy^uch^ wSwas centered a .he University of Edinburgh. d.e UK was r=^ j^^d-ng 
*e field al this time. However, in the early 197ffs a high-level UK e<»|"''?~- ""^^ ^ 
Ughthill. turned down AI for a large grant The UK. ai the ume was P^Jf'"* " ^ 
toded aider the tide, "National Development IniUatives". This largely J^couraged^ W AI po.q>. 
^Tof^m subsequenUy came to *e U.S. S««E feigenbaum »dPMcCortuc^^^^^ 
Gentration 175-176. Also see M. Minsky,''The Problems ami the Promise, m P. Wmstonano 
^S^l edi TA* M Business. MTT Press. 1984, p. 246. Re^n^y- 

•M^^^^L in information sciences has included a sizeable component of AL Ir^ormaaon 
Technology R&D, OTA, U.S. Government Prinnng Office, 1985. 

1 1 E. Feigenbaum and P. McCorduck, The Fifth Generation. Addisson-Wesley. 1983, p. 65. AO 457 of 
3/63 Heuristic Programming. 

12 C. Green. "AI During IPTOs Middle Years," in T. Bartee. ed.. Expert Systems and Artificial 
Intelligence. Howard Sams, 1988, ibid., p. 238. 

13 rA<S«e<fao/Arii/!cia//«eH/««/ice,NauonalInstiiuies of Health. FO-2071. 1980. pp. 18-19. 

15 ^'xpandtag AI Research and Founding ARPANET." by U Roberts, in Bartee, ibid., p. 229. AO 1058 
of 7/67 for "Inlelligent Automata." 


as knowledge representation, problem solving, and natural language structure, and to 
applications in areas such as expert systems, automatic programming, robotics and 
computer vision.^^ This AI research was carried out mainly at MIT, Stanford, Stanford 
Research Institute (SRI), Bolt Beranek and Newman (BB&N), and later Carnegie-Mellon 
University (CMU). which have remained major AI centers to date. However, C. Green, 
who was in charge of this early AI work at ARPA. felt that there was more money than 
good ideas at the time.^'^ 

In the early 1970s the early developments of ARPANET already expanded the 
range of possibiUties for interactive computing. At this rime another NIH-supported AI 
effort was started at Rutgers focussed on problem solving.^' This and other NIH AI- 
reiated medical research resource development programs quickly took advantage of 
ARPANET wherever possible, together with other networks, to speed up exchange of 
research information,^^ 

The Xerox Palo Alto Research Center (PARC) was set up near Stanford in the early 
1970's by R. Taylor, who had been director of ARPA's IPTO. One of the earUest efforts 
supported there by ARPA was the development of a widely used version of LISP, "Inter 
LISP." Other LISP "dialects" began to proliferate, and were eventually coordinated in the 
late 1970's by meetings and ARPANET teleconferences promoted by DARPA.^^ 

In the early 1970s there were proposals to construct a new computer especially 
configured to execute LISP. ARPA, apparenUy, did not support these efforts expUcitly, 
partly because of the IPTO experience with ILLIAC IV.22 There were also concerns at the 
time about government support of computer building outside of industry, with "cheap 
labor" of graduate students.23 MIT persisted, however, and in 1980 LISP machines had 
been constructed and used in MITs Laboratory for Computer Science (LCS), and Xerox's 
PARC, which had built its own. and were offered for sale by companies formed by ex- 

1 The Early Yeais, Founding IPTO, by JCR Licklider, in Bartec, ibid., p. 220. 
1 "A.I. During IPTO's Middle Yean," by G. Green, in Bartec. ibid., p. 237. 

1 8 Interestingiy. ARPA^fET was not greeted enthusiasticaUy by ail members of the AI community, cf. 
Roberts, ibid. 

1 9 S. Amarel. "Problem Solving," Chapter 4 in T. Banee. cd.. Expert Systems, ibid. 

20 Seeds of Artificial Intelligence, ibid., p. 69. See also Tompuier Networl« - Prospects for Scientists." 
by Alien G. Newell and Robert F. Sproull, Science, Vol. 215, 1982. p. 851. 

2 1 Footnote by R. Engelmore in Bartee ibid., p. 244. 

22 Roberts, ibid., p. 232-3. 

23 Discussion with M. Denicoff, 6/89. 


MIT researchers. Many of these LISP computers were subsequenUy purchased by AI 
researchers with ARPA support, and by other government laboratory groups.^^ The 
computers involved in a typical current AI laboratory (NRL) are shown in Fig 1. RecenUy, 
however, LISP execution on the CRAY (general purpose) supercomputer, in a test 
supported by DARPA, has been demonstrated to be faster than specialized LISP 

1 . Applications 

In the early 1970's ARPA's first major concentrated AI applications project was 
begun as pan of an interdiscipHnary effort toward the Speech Understanding Research 
Project ( SUR). This was the first large effort on computer speech, and it was undertaken 
despite a National Academy of Science Committee's (Pierce Committee) negative 
recommendation. At the same time there were also some encouraging developments, such 
as a device to automatically generate phonemes from speech 26 A very strong motivation 
for this program was the great advantages that were envisioned of being able to 
communicate with computers with speech. 

The ARPA SUR program was initially planned to have two 5-year phases, with the 
first having the goal of a 1000-word vocabulary, uttered by a limited number of speakers in 
a relatively quiet room.27 Some AI researchers, however, regarded such quantitative goal- 
setting as premature at that early stage of AI research. The SUR project funded several 
competitive approaches and there was also a broad supporting research program. The 
following summarizes the results of the first phase of this programr^s 

Initially the LISP machines were specialized mainframe computers. Uter, with the increase of poww 
of sm^ier machines. LISP could be executed with interactive graphics on personal computers, and 
more recendy, on a single chip. 

25 IEEE Spectrum, 1989. 

26 Roberts ibid., p.234. AO 1943 of 8^1. 

27 Green, ibid., recounts that Roberts. IPTO head at the lime, said he wanted 10^ words, ^d if not that, 
as much as could be done. However, a committee of peers was set up by ARPA. and decided 10- 
words was a reasonable goal. 

28 R S Enelemore et al., "Hearsay - n," in R. Englemorc and T. Morgan, eds.. Blackboard Systems, 
Addifon-W^^;. 19^^^ p. 25. L. E^man, et ah. "The Hearsay-II Speech Understandmg System: 
Intei^ng Sledge .o Resolve Uncertainty." in R. Englemore and T. Morgan. Blackboard Systems, 
ibid., pp. 60 • 75, compares the competing systems. 



16 Node 

Figure 1. NRL's NCAR Al Computing Facility 

Three organizations finally demonstrated systems at the conclusicm of the 
project in 1976. These were Carnegie-Mellon University (CMU) that 
actually demonstrated two systems; Bolt, Beranek and Newman (Bbrs), 
and System Development Corporation with Stanford Research Institute 
(SDC/SRI)....The system that came the closest to satisfying the pngmal 
project goals was the CMU HARPY system. The relatively high 
performance.. .was largely achieved through 'h"^-winng 
information...into the system's knowledge base. Although HARFY made 
some interesting contributions, its dependence on extensive pre-knowiedge 
limited the applicability of the approach to other signal-understanding tasks. 

The second phase of SXJR, however, was not carried out Some feel this was 
because the first phase did not produce a sufficiently impressive product.29 However, 
performance was recognized to have been limited, in part, by the speed of the available 
computers, and some improvements would await a new generation of computers, several 
years away. During the SUR project there were a number of proposals to construct LISP 
computers, motivated by the expected advantages for speech recognition, but as mentioned 
above, these were not supported by ARPA. In order to get an objective assessment and not 
lose track of SXJR research achievements, a small effort was supported by ARPA and ONR 
to review and document the SUR effort^ 

Besides leading to a number of specific research contributions to the field, 
summarized in Fig. 2, the SUR effon developed methods that have had wider application. 
One such spinoff is tiie "blackboard" technique, which was a feature of a second SUR 
system developed by CNOJ, Hearsay-H. This is an approach "for coping with problems 
characterized by the need to deal with uncertain data, make use of uncertain knowledge, 
and apply a nondeterministic solution stiategy."3i AppUcations of this technique include 
image recognition, signal understanding, protein-crystallographic analysis, and data 
fusion.32 The blackboard techniques developed under Hearsay-H were adopted as the 
framework for the ARPA- sponsored HASP program on ocean surveillance signal 

29 Licklider. ibid., p. 226. 

30 -Review of the ARPA SUR Project" ONR report by Wayne Lea and June Slwup. S^ch 
Communication Research Laboratory, January 1979. and " AI Developmem and the Office of Naval 
Research,** by M. Denicoff. in Bartee, ibid., p. 280. 

3 1 R. Englemoie and T. Morgan, Blackboard Systems, Addison-Weslcy, 1988, p. ix, 

32 Ibid Sec also Computer/Vision, by D.H. Ballard and C. Brown, Prentice Hall, 1986, p. 505. 

33 H. P. Nii, et al.. *'Signal-to-Symbol Transformauon: HASP/SIAP Case Study,", in R. Englemore and 
T. Morgan, ibid., pp. 1235-1236. 


i: II 10 •» 3 7 4 S i 3 2 I 

• jst TsiE nrra co*u (c«i,.; 

. MCvKtcy (>9CS) 

. Cdattnuatta $»««eft C'lQO $«ae«ae«*) 

. L«r|« 7«eabwl«C7 (>1000 «Mrda) 

. Cf{Mtl«« Um at LlRfMisciA Coil«r«lB« 

. Spaa4 («100«:?SS) 

. Inpvc tavtroKBwie (Tanuua toom MIm). 

. »Mict9l* SpwMrt, 7utM«bUicy 


• . KarpT JCC) 

. tacc^rsecd «l«o«ck 

. Bm«»Mrfe Syaco 

. HCAUAf tcs) 

. ta«n«irt«ne bwwlatfg* Sowcas 

. wet (»W> 

, Oalfon ScorlBf Proe^dur* 

. UfieUac Maltslbis Stracaiy • 

. Liacoia Utoracory S««c« 

. ktXAlSAX t $y«e«a (CS) 

. Ora(«a S*acm (Ofll) 

. »t S»acaa 

. srncNLZS (iwi • 

. SBC vtm 

. SOC IJJ» Sya« 

nsTct oojitiat ua suutai TtoKoois 

. Utt>ltl(nt Uam Saareh »/o B«cltcraeklB« (CO) 

. iMUad CTlTlBt C«. SK) ■ 

. PT»«a«Ul»t£ Uof Itlwllhpod) Scarlai 

, racearaa lawladsa IU*rt««acaclaaa 

• rmxiD uxwuisttc coJismDrcs amo ststw ratanwiCL 

. D««aiapMac «i y^aswraa «( CwplaxlCT (CO) 

. £f(ceca 9( SraBchlfti Factor oa P«r<a««aaea 

. Cffaeta ot ConXuaabUity la V«ea»«i«rT iOUl 

, Ufaeta «< Sanetaaa Ua^ch oa Parfswaea (SDCi 

, Ufaaca o( Voeaaularr «« Parforaaoea 

. SaMaUe « -Jart va niaaacic Aeoiracr..- 

, AbUclos Stwdtaa: oi Sfnux aad $««aetca (Cm; 


, Isprovod Asaaactc P»»a«a«ar Ucraecera 

. Phsntcle AMirsta Taeftalauoa • 

. Allaphaatc TaepLacaa (C«IJ 

. Ptonatte Uttlca OBJl) 

. Phoaacic Sagsaticaciaa «ad LaballRf Ma«l»o4» 

. Phaaolotlcai «ul«a and UzIcaI Ucrlovai 

. CoBVllia« «ed Taadnf Fhomolafleai laiaa 

. Laslcal Oaaaatni Saeworh (SBH) 

. Uard Juaewt Xuiaa (CflJ. iu»pp)««»««»»»««» 

. Uavd Varlf icadoa (BSM. SOC, Cfll) 

. Syaeax and Psraiaf 

. Pacatflt Srrarfttt Seclnta 

. Arbtetm iMttiat Polnea (BU. OBI) 

. Subacvma chac Aran't !ioBcat«fMla 

. Pattanasca CraMara (S«).... rrriii"lit'y 

. PraiMCiu (BUeettraa, tuk. tUav CaaacraUcat ttX. SDC>, 

BtPEXufcrrAi. xnscAACS 

. Preaedic Ald» Spaaeh taeagaielafl 

. S«LLabl# Oataatlaa 4na Cm ta Uca«Bittaa. 
. Spaeeresraa badlai (tUalitaa. SBSI, Cflt)... 

• Spa«eh QACdbaaaa 

. Tranacrtpctoa ?roead«raa 


Figure 2. Review off Project SUR Research Developments 

Source: W. Lea at J. Shoup. "Review of the ARPA SUR Project." ONR Report 
January 1979. 


The HASP program began in 1972 as an effort to use AI techniques to 
automatically recognize signals from seismograms from underground explosions or in 
sonagrams used in ASW.34 HASP was to use the ELLIAC IV, the most powerful 
computer at the time which was being exploited for seismic underwater acoustic research. 
HASP and its successor program, SIAP, showed some success, but the effort was not 
considered worth continuing at the time.^^ 

Also stemming from the SUR work are the linear predictive codes later used in the 
Morse Code reader effon by MIT's Laboratory for Computer Science,36 discussed 
separately in Ch^ter XXH; and in secure speech systems used by the military. SUR^ 
generated technology has also had an impact on voice recognition used in miUtary training 
systems, such as TRIO, developed in 1983 for radar intercept opcrators.37 in the late 
1970's, IBM began research on speech recognition, partiy building on the SUR results, 
and adding some new approaches.^^ 

Dr. G. Heilmeier, upon becoming DARPA Director in 1975, raised "very 
fundamental and pragmatic questions about the AI research field."^' Heilmeier says.^ 

I tried to apply my catechism questions: What are the limitations of cunent 
practice? What is the current state of technology? What is new about these 
ideas'' What would be the measure of success? What are tiie milestones 
and the "mid-term" exams? How wiU I know you are making progress? I 
asked these of all the programs, but for AI I didn't get any answers. This 
sent the AI community into turmoil apparentiy no one had challenged 
them in the past 

"It wasn't tiiat I was never a bcUever in AI, I just wanted tiiem (the AI program 
leaders in IPTO) to answer basic questions, and they couldn't"*i Heilmeier recounts that 
he "saw no investment strategy - this was the ultimate in laisscz faire research." The AI 

3^ Ibid., describes the HASP and foUow-on SIAP projects. 

35 "Later Years at XPTO," by R. Kahn, in Bartee. ibid., p. 248. HP. Nii, et al.. "Signal-to-Symbol 
Transformation...." ibid., discusses analyses by the MTTRE Corporation of experiments companng the 
perfomance of SIAP with expert sonar analysts. Also, discussion with H. Auiand. 3/8». 

3^ Discussion with Mr. A. Vezza. 4/89. 

3 7 BB&N, Science Development Program, Annual Report 1988. 

38 Lea and Shoup, ibid., p. 30. 

39 R. Kahn, in Bartee, ibid., p. 246. 

*0 Interview with Dr. George Heilmeier, 8/29/89. 
41 Ibid. 


researchers, in his view, wanted "a cashiers booth set up in the Pentagon-give us the 
money and trust us." The essential issue in Heilmeier's mind was one on "faith versus 
accountabiUty." The perspective that he was given was that Al researchers were too busy 
to write proposals or even to write papers on their research. Moreover, AI was too 
complex and difficult to explain to non-experts. Energized by this challenge, Heilmeier 
reviewed the AI researchers' ARPA proposals and their research material ("Apparently I 
was the first ARPA Director to read their proposals. ")*2 He concluded that the AI program 
was insufficiendy structured and focussed to justify the level of funding and attention diat it 
had been receiving. 

Not receiving a satisfactory answer. Dr. Heilmeier asked the JASONs^^ to look at 
the AI program and got "a lukewarm endorsemcnL"^ Heilmeier's solution was to specify 
some mUitary appUcations where AI could be appUed and focus a major portion of the 
DARPA program on these. The result of his review was a niajor shift in the balance of 
work toward appUcations.-^ Heilmeier identified several specific appUcations programs for 
AI, notably the ACCAT (Advanced Command and Control Applications Testbed), and the 
automatic Morse Code reader at MIT,^« The total AI budget did not go down under 
Heilmeier, but die balance between fundamental and appUed definitely shifted. 

There were misgivings in the community (and still are) about expecting too much 
too soon from AI without sufficient research foundation. Heilmeier contends that his 
focussing on applications supported the development of the technology and that he 
recognized the need to provide continued funding for basic research. However, he made it 
very clear that continued funding of basic research was contingent on the conduct of 
appUcations work as well. 

42 Ibid. 

43 jASONs are a group of leading U.S. physical scientists who devote their attention to problems of 
science and national security. The JASONs (named after Jason of Greek mythology) were organized 
originally in 1960 at the Institute for Defense Analyses with the support of the then Director for 
Sfcnse Research and Engineering. Dr. Herbert F. York. Sec, H.F. York, Mahng Weapons, Talhng 
Peace, Basic Books, New York, 1987. p. 153. 

44 Heilmeier, ibid. DDR&E also asked an external review group to assess the DARPA AI Pjo^sin 
the 1970s; their conclusions were paraUel to Heilmeier's communication firom Dr. A. Flax. IDA, ZAW. 

45 This change is discussed by LickUdcr and Kahn. IPTO directors at the time, in Bariee's. p. 225 and 
p. 246.. 

46 Heilmeier says he also pushed two other application areas, ASW signal understanding (HASP) and 
image understanding. Sec also Kahn, ibid., p. 247. 


In the command and control areas DARPA believed that not only was the 
technology that had been developed in the AI and ARPANET networking programs far in 
advance of what was available to the Services, but also that this technology could solve 
existing problems including, importantly, those due to widely differing computers and the 
management of distributed files at different locations. ACCAT was set up as a joint 
DARPA-Navy effort towards embodiment and test of many of these technologies, 
including management of distributed, relational data bases, RITA for file query 
management, and the LADDER natural language system, in a controlled laboratory 
environment at NOSC. The ACCAT simulated a Navy command center and would 
communicate via networking with other command centers and data and computer 
resources.*^ ACCAT also provided additional capabiUties for war games played by the 
Pacific fleet Changes in the existing ARPANET technology were also required for 
ACCAT to interface with "MIL spec" computers. ACCAT was also a test bed for 
developing and testing approaches to a secure network environment, since several data 
sources in classified facilities were Unked together with unclassified nodes of 
ARPANET.^ Chapter XXmfiather reviews the ACCAT project 

Another response to the DARPA push toward more AI applications was a project at 
MIT's Laboratory of Computer Science (LCS) to design and construct an automatic 
translator for manually generated Morse Code, using AI expert system techniques. 
Building on previous work at the Lincoln Laboratory, and some of the results of the SUR 
project. AI techniques were applied to the interpretation of somewhat garbled and 
incomplete word streams and brief introductory transmissions from acmal Morse Code 
tapes to make a "best" translation. The Morse Code project was considered successful by 
MIT and the results were communicated in the late 1970's to U.S. government laboratory 
groups. The National Security Agency considered the results sufficiently promising to 
continue making further improvements toward practical applications.'*^ Chapter XXH 
elaborates on the Morse Code Project 

47 Discussion with D. Small. NOSC 3/89 with R. Brumderberg. 6/89. Cf. also an article in 7. Oe/cnj« 
ResearcK "ACCAT: A Testbed for Exploring Change." by F.H. HoUisier. Special Issue 78-1 on 
Tactical Command and Control, 1978, p. 39. 

48 "ACCAT and FORSCOM Guard Systems." by M. Soleglad, address at the 4th Seminar on DoD 
Computer Security Initiative, Aug. 1981. 

49 While the MIT Morse Code effort went on for nearly four years the main '^^^^^^^^^^^^^ 
available by the second year and the government laboratory smipUHcaDons and improvements were 
made after that Discussion with Dr. S. Squires, May 1989. 


Other defense appUcations of AI have been pursued based on the work initiated by 
ARPA. In the late 1970*s a system for planning Air Forces missions, Knowledge-Based 
Systems (KNOBS), was developed by MITRE widi Air Force suppon, and tested on 
DARPA supponed computers at project MAC. Later, a similar planning system. 
Knowledge-Based EngUsh Entry Crew Activity Planner (K>JEECAP), was developed by 
NASA for use with the space shutUe.^o Late in the 1980's, the SDI batde nianagement 
program began to construct a test bed faciHty which incorporates many of the advances in 
computers, software, and AI pioneered in the DARPA program,'^ 

2. Commercial Developments 

In the late 1970s, perhaps stimulated to some extent by the new DARPA emphasis 
on applications, and encouraged by the success achieved in DENDRAL, a number of 
expert or knowledge-based systems began to be developed for applications. These 
applications have been developed mostly in industry and many by individuals whose 
training in AI technology was supponed by DARPA. Some AI application systems which 
appear to have reached the most advanced stage of conamercialization include: DECs R-1 
or XCON for designing computer circuits; the DIPMETER ADVISOR for oU well logging 
data analysis, by Schlumbcrger. the ACE line fault diagnosis program by AT&T, the 
EXPLORER geological exploration program by SRI, and the STEAMER computer-aided 
instruction systems for Navy engine-room personnel, by BB&N.52 A recent review listed 
approximately 150 expert systems in use.^^ 

Several companies sprang up to supply expert system assistance in areas such as 
financial investment, information services, and computer circuit design.34 By the late 
1970s some ten companies in the AI software and hardware areas had spun off from the 
MTT AI group alone.55 A handbook of AI, supported by DARPA and NIH, was pubUshed 
by Feigenbaum.56 Robotics-type activity in industry increased considerably in the late 

50 "AppUcations 1 - Space," by Edward L. Lafferty. in Bartee, Arid., p. 9, and discussion on June 1989. 

5 1 "Computer Aided Beuer Management" by D. Dalun and Y. Smith. Aerospace America, June 1989. 
p. 40. 

52 "Amplifying Expertise with Expert Systems." by R. Davis in Winston, ibid., p. 188. 

53 E. Feigenbaum. P. McCorduck. and H.P. Nii. The Rise of the Expert Company, Times Books, 1988. 

54 "Artificial InieUigencc is Here," Cover story. Business Week, July 9. 1984. 

55 "Project MAC," ibid., foldout. 

5 6 "Seeds of Artificial InteUigence," ibid., p. 63. A later encyclopedia was edited by Shapiro. 


1970s 57 While there was earlier IPTO interest, higher level decisions at DARPA were not 
to emphasize robotics, at that time, although it was one of the main areas of interest of the 
MIT and Stanford AI groups still supported by IPTO 58 Later, the DARPA IPTO program 
included substantial robotics support, including the recent Strategic Computing program 
effort towards an autonomous land vehicle. 

An important impems to the application of AI in industry occurred with the 
appointment of former DARPA Director, G. Heilmeier, as the Senior Vice President and 
Chief Technical Officer of Texas Instruments (Tl). Under his direction TI became one of 
the first major companies to embrace AI as a central business thrust.^^ Today, TI is 
regarded as the leading AI company with its products, including its Explorer Lisp machine, 
an expen system shell. Personal Consultant, custom expen system for industrial and 
military appUcations.^o Heilmeier's predecessor as DARPA Director, Dr, Steven Lukasik, 
as Corporate Vice-President for Research at Nonhrup Corporation, supported the 
development of an expert system manufacmring process planner for internal use.^i More 
recenUy, IBM, GE, DEC and other larger companies have shown some interest in AI.62 A 
recent estimate is that the commercial AI market is approximately $600 million today, 
growing from about $20 million in 1983.^^ 

DARPA AI support also contributed to development of several aspects of computer- 
aided instruction (CAI). Many of those active in CAI and AI were very interested in the 
prospects of an intcliigcni computer systcnas for education and training. An MIT AI group 
under S. Papert made a major contribution in writing a LISP program for LOGO during 
project MAC in I960." LOGO was used in many elementary school experiments, and 
improvements were supported eventuaUy by NSF and the U.S. Department of Education. 

57 A review is given by J. Michael Brady in Winston, ibid, p. 179. and a brief historical review is given 
in Robotics by K.S. F., et al.. McGraw Hill 1987, p. 4. 

58 Perspectives on early robotics iniiiaUves at ARPA and ONR are given by Bartee, ibid., by Roberts, p. 
231 and DenicofF. p. 298. 

59 E Feigenbaum, P. McCorduck, and H.P. Nu, The Rise of the Expert Company, Times Books. 1988. 
DP. 174-188. describe HeUmeier's leading role in advocaung AI development as a business thnist for 
Texas Instrument Heilmeier's activity, at DARPA and Texas Insuuments icgardmg AI also is 
discussed by Licklider and Kahn, in Bartee, ibid. 

6 0 E. Feigenbaum, P. McCorduck and H. Nii. ibid. 

61 ibid., pp. 24-30. 

62 Business Week, ihi± 

63 H Ullman, "Machine Dreams: Future Shock for Fun and Profit (Failure of Artificial Intelligence lo 
Meet Expectations)." New Republic, Vol. 201, July 17, 1989, pp.12-13. 

64 Information Technology R&D, OTA. ibid., p. 160. 


In 1980, LOGO was implemented on microconiputers and in 1982 a company, LOGO 
Computer Systems Inc. was formed by some of the MIT group to supply a growing market 
for LOGO diskettes.^ 

Another Al-based computer-aided instruction tool was STEAMER, developed by 
BB&N for the Navy to teach ship engine-room procedures. STEAMER was, apparentiy, 
an outgrowth of SOPHIE, an intelligent circuit analysis program, in turn based on a 
University of California (Berkeley) circuit analysis program, SPICE, which had been 
supponed by DARPA.^^ SOPHIE was regarded as one of the first "Intelligent Computer- 
Aided Instruction" (ICAI) programs and led also to several militaiy training programs such 
as QUEST for troubleshooting.^^ 

In general, the relation between AI and CAI seems to be paced by progress in the 
fundamental AI area of knowledge representation. Some feel the interaction has benefited 
AI more than the other way around.^* D ARPA-supported AI efforts on low-cost computer 
imaging, combined witii results of its networking programs, particularly by satellite 
between widely supported areas were essential to the development of SIMNET. now being 
used by the U.S. Army to simultaneously train tank crews in the U.S. and Europe in 
batdefield tactics.^ 

3 . DARPA Strategic Computing Program 

In 1983, DARPA commenced its Strategic Computing Program, challenging advances in 
computer technology and AI applications."^® This program approximately quadrupled 
annual Federal funding of AI and related hardware R&DJ^ Three specific AI application 
areas are featured in this program: (1) A "pilots associate." incorporating natural language 
interactions witii computers and expert systems to monitor vehicle performance and 
control, and generate alerting statements, giving new impetus to speech recognition 

Project MAC 25Ui Anniversary, ibid., foldout. 

6^ Targeting the Computer, by K. Flamm. Brooking 1987, p. 69. 
QUEST was developed by BB&N in 1986. BB&N, ibid., p. 46, 

fi8 In the late 1960's and early 1970's one of the greatest impacts of the advances in AI was on the field of 
psychology. Together with the intensified study of activity of the neural system and the processes 
involved in perception, AI opened up the field of cognitive psychology. This has had considerable 
influence and interaction with efforts to automate military training and testing. D. Flcichcr. ibid. 

^9 SIMNET was firei demonstrated in 1987. BB&N and Information Technology R&D, OTA, ibid. 

''^ Strategic Computing Program, Annual Reports, DARPA. 

7 1 Ir^ormadon Technology R&D, OTA. ibid., p. 96. 


research; (2) Naval battle management, again involving natural language interfaces to 
access and query extensive data bases, together with graphics, integrating fleet status 
information and decision aids, (reminiscent of some of the work stated in ACCAT); and 
(3) robotic autonomous land vehicles, emphasizing computer image-comprehending 
systems. After extensive preliminary development and trial, systems of each of these three 
types have advanced to prototype stages and part of at least one (fleet status) is undergoing 
Service evaluation.72 Along with these specific projects, a supporting research program is 
going on to provide needed developments in microcircuits and information processing 
techniques, together with opportunity for access to all these developments by research 
workers- Each of these projects involves the most advanced and powerful computers that 
can be consmiaed and still be compatible with the respective operating conditions. 


The major push for the development of Artificial IntelHgence can be credited to 
ARPA's funding in the late 1960's and early 1970's. The interplay and interaction of AI 
with computer development in this early period was very broad and strong. The needs of 
AI research for interactive progranming were a major factor motivating support for the 
development of computer time sharing, and for the "user-friendly" characteristics of 
computers, which have become major characteristics of tiie personal computer today. At 
the same time. AI's developments were paced by the great improvements in computer 
hardware capability and the fall in costs of computing. 

The impact of AI on related sciences, such as cognitive psychology, has been very 
great^ The interplay of AI with Computer-aided instructions also has been considerable. 
The first ARPA attempt toward AI appUcation in this early period, the Speech 
Understanding project (SUR). was motivated by its very high potential payoff for 
enhancing human-computer interaction. The SUR results, while useful for further work, 
indicated the expectations at the time had been too high for the existing computer 

By the mid 1970s, various specific AI applications began to appear. Perhaps the 
most important of these was the DENDRAL expert system, which was developed as a joint 
effort between some of the Stanford AI group, who had earlier ARPA support for 

72 Being essenuaily software, it may be possible to test parts of the fleet batUe management system 
separately on existing computer equipmcnt- 

73 cf. Margaret Bodan. "Artificial InteUigence in Psychotogy," MfT Press, 1989. 


"heuristics" which were the basis for DENDRAL, and medical researchers. NIH suppon 
was responsible for carrying DENDRAL through a long period of experimentation to 
success. WhUe ARPA maintained some support to DENDRAL throughout the 1970s, the 
role of NIH in supponing knowledge-based expert systems as demonstrated in medical 
applications was instrumental in the visibility of AlJ^ 

Greater emphasis by ARPA toward applications in the mid 1970s led to accelerated 
AI developments in a number of specific areas. Part of the ARPA push derived from an 
appreciation that AI, with its own great problems of software development, might be able 
to improve the efficiency and lower the costs of software production, which was beginning 
to appear as a major economic factor in computer use. The results of this period of ARPA 
AI support seems to have met this goal, to some extent After an initial deUy. probably due 
to the ILLIAC IV experience, ARPA funded the LISP machine development at MIT. AI 
researchers have designed relatively inexpensive LISP computers. Now a commercial 
item, these are powerful tools for complex software development and used widely by 
industry and in government laboratories. Corresponding advances in "intelUgent" terminals 
also have been made. 

On the other hand, this ARPA applications emphasis has, in the opinion of some AI 
researchers, retarded programs on more fundamental and difficult problems which underiie 
the capabiUties of all appUcations. Today, opinion seems to favor the view that progress in 
the AI applications area in the near future wUl occur by use of existing Al-related 
technology in well-defined areas. The majority of miUtary appUcations, for example, 
seems to be occurring in the use of expert systems in "smart weapons." planning. CH data 
fusion, repair practices, and training.^^ 

The DARPA Director G. Heilmeier's effort to force "top down" AI appUcations in 
the late 1970s seems to have been partly successful. The Morse Code Reader, a relatively 
easy problem compared to speech recognition, transitioned quickly to a laboratory user 
group in NSA. ACCAT, which pushed a variety of AI technologies, perhaps too hard, 
within a rather diffuse C^ training environment, had Uttle direct impact, but did solve some 
related communications problems and whetted appetites for what might come later. 
HeUmeier's view is that ACCAT succeeded in changing the view of C3 in the military: for 

S, Amarel, "Current AI Research," in T. Bartee, ed., ibid., p. 259. 


See R. P, Bonnasso, "Maitary Systems," aapter 7, in T. Bartee, ed.. Expert Systems , ibid . and S. 
Andriole. "Artificial Intelligence and National Defense," Chapter 19 in S. Andnolc, e(L. Applications 
in Artificial Intelligence, PetrocelU Books. Inc., 1983. 


the first time C3 was approached from an information management perspective integrating 
decision aids» AI, and information management technology^^ 

One important outcome of diis turbulent DARPA AI period has been a very efficient 
technology transfer to the commeiciai sector. The first major industrial application of AI 
was made in the oU prospecting area, by Schlumberger. This drew broadly, like the other 
appHcations in the same period, on the AI technology being developed largely with ARPA 
support. Much of the development of commercial AI has been spun off from university 
research programs, chiefly at MIT, Stanford, and Camegie-MeUon, supported by ARPA. 
Several key players in ARPA's IPTO AI program have gone into the commercial sector, 
while others now are pursuing academic research in AI. 

Dr. Hcilmeicr. who was highly skeptical of AI program in IPTO when he arrived, 
subsequendy went to Texas Instruments, where there is now an AI applications thrust with 
an emphasis on symboUc processing and object oriented computing.'^ He sees "symbolic 
processing as the future of computer appUcations." He stated that for TI commercial AI 
applications are foremost; AI has permeated the commercial sector too a much greater 
degree than the miUtary. A problem he noted, based on his experience witii such projects 
as ACCAT and HASP, was a reluctance of potential military users to adopt "revolutionary" 
processes. Thus, he felt tiiat it might be another ten years before widespread application of 
AI in miUtaiy systems.78 However, there already have been some identifiable miUtary AI 
appUcations, such as TI's advanced USP processing chip for "sman" missiles. 

In reviewing the AI program at ARPA, it is important to recognize that the field 
itself was in its infancy when ARPA began its support The overall vision of licklider Mid 
his successors was to enhance the ability of computers to perform in intelligent ways witii 
an underlying premise that such improvements would be important to defense applications. 
Reflecting on die impact of this program, Robert Kahn. a fonner Director of IPTO, noted'' 

The main impact of AI to date has been to broaden die tiiinking of some of 
the research and operational people in Defense, and to make them aware that 
they can do more witii electronics tiian just some of die programmed kinds 
of things tiiey were used to in the past • tiiat inteUigence in these systems is 
defiiutely a possibility in the future. 

"^^ Heilmdcr, interview. 8/89. 

77 Ibid. 

78 Ibid. 

79 Kahn, ibid., p. 252. 


In one sense, AI hasn't really made an operational impact yet because there 
are no embedded AI systems in operation, and the policy for supponmg 
them is not there. A few experimental systems are bemg used and 
evaluated; however, AI technology has had a significant impact on some 
contractors who can now develop software more effectively. It has also 
enlightened a lot of people through concrete demonstranons of what the 
technology can do - 

DARPA's Strategic Computing program, begun in 1983, can be looked on as an 
attempt to bring AI and computer technology together, with a focus once more in several 
appUcations areas. Some of the Strategic Computer objectives revisit, in a more mature 
fashion and with much improved technology, previous attempts in the speech recognition 
and C^ applications. 

Recently, with the increased interest in parallel structures to achieve faster 
computing, the analogy to research systems has been rediscovered, with munial benefit to 
computer architecture, to cognitive studies and AI. DARPA outlays for AI up to inception 
of tiie Strategic Computing program from project records appears to be about $120 million. 
A recent estimate of die value of the commercial market is about $600 An 
increasing number of military systems are planned to incorporate AI in a more or less 
essential way (see Fig. 3b). Expenditures on these systems are estimated as several billion 

80 Cf.Ref,63. 


Systems that could incorporate KBS: 



OH-58D (AHIP) 













CVN 71/72/73 







V-22 (JVX) 














Figure 3A. Major Defense Acquisition Programs That Could Incorporate 
"Knowledge Based Systems (KBS) 

Systems that will incorporate KBS: 













8-1 B 



Figure 3b. Major Defense Acquisition Programs That Will 
Incorporate "Knowledge Based Systems" (KBS) 
























▼ K 

mm h» mm mm S 'JL 




A.L yTO 




. IMS 
™ ^ SHARING mm 




W.E. Co. 
















- r 






SPACE ^ ▼ 





^ •" [••■•••* UNIX 



I > 
1 I 



I ^ SVI 

!▼ I 

■<DDt C 





LMI 1977 



I ▼ 













The Morse Code project was undertaken by MIT's Laboratory for Computer 
Science in the period 1974-78 in response to an ARPA request to look into the problem of 
replacing a human high-frequency radio operator interpreting manually-generated Morse 
Code with an "inieUigent" computer system. Using available AI techniques, a successful 
automatic "Morse Code reader" was developed by the MIT group and picked up quickly by 


For many years a substantial ftaction of radio traffic in the high-frequency spectrum 
involved manually-generated Morse Code. These signals were generally characterized by 
many irregularities, notably in duration of the long pulses (dashes) and spaces between 
short (dot) pulses, in which individual "senders" often had distinctive patterns. 

The problems of "reading" Morse Code is made more difficult by frequent 
interference of other signals and the characteristic "fading" of high frequency radio 
transmissions. On the other hand, the patterns in these simations and in the message 
protocols and language of amateur radio all seem to be used to advantage by experienced 
radio operators. Recently, most Morse Code transmissions have become "machine" or 
computer generated, with far less irregularity and so much easier to translate automatically. 
There arc commercially available systems to carry out this function, i 

As part of an effort to steer the ARPA AI program more towards applications,^ 
Dr. G. Heilmeicr, ARPA Director in the mid 1970's, generated a list of military problem 
areas where he felt AI might be helpful. One of these problems, apparendy from NSA, 

1 Gary L. Dexicr, Shortwave Radio Listening With the Experts. H. Sams Company. 1986. p. 325. 

2 R. Kahn. p. 246 in "Expert Systems and Artificial Intelligence." T.C. Bartce, Ed., H. Sams, 1988. in 
Bailee's book. 


was that of "reading" manual Morse Code traffic ^ Responding to Heilmeicr's pressure! 
J.C.R. Licklider, then head of ARPA's IPTO, called A. Vezza of MTTs Laboratory for 
Computer Sciences (LCS) to ask if they might be able to do something on this problem:^ 
(LickUder had just come back to IPTO from the LCS.) Besides being quite famiUar with 
the MIT's LCS generally, he had been a collaborator with A. Vezza in the LCS 
programming technology group. The major occupation of this group previously had been 
with development of automatic programming technology. 

This was, acmally, the "second time around" on this problem. In the mid to late 
1950's MTTs Lincoln Laboratory had developed MAUDE, which was a computer program 
to "map" Morse Code symbols into alphabetic and numeric character sets.^ MAUDE used 
some rudimentary "rules" in this mapping, some statistical and others including the 
maximum number of dots and dashes in a legitimate Morse Code character, and dealing 
with "pairing" of symbols which are often confused. NS A attempted to apply MAUDE to 
manual Morse but found this impractical.* In contrast to machine-generated Morse which 
was quite easily handled, NSA resigned itself for many years to the view that manual 
Morse required a human interpreter. 

After UckUder's request, Vezza spent about three months reexamining the MAUDE 
results and thinking about the problem. Vezza concluded that the AI tools and the 
improvements in computing power then available could lead to a solution. No break- 
through seemed to be necessary, and so the MIT AI group, mainly concerned m\h new AI 
developments, was not involved. Vezza envisioned that AI "expert" techniques could map 
die inegular Morse Code streams not just into characters but onto sets of words taken from 
stored vocabularies, with corrections for grammatical structure. Compared with th^ 
difficuU AI problems of translating namral language, the MCR problem was much simple;^ 
a "toy".'' Further, the problem had been discussed witii LCS staff, some of whom^ wer6'>^ 
amateur radio "hams" and there was much enthusiasm for the notion of constructing an 
"artificial ham."^ In fact, the LCS group began to set up such a "ham" station on the roof 
of the LCS building. However, the- FCC pointed out the possible iUegaUty of copying 

3 Testimony of Dr. G. Hcilmeier, p. 4908 in Hearings on Military Posmie before Committee on Armed 
Services, DoD authorization for 1976. and 76T. H.OJI. 94th Congress. 1st Session. Pan 4. 

^ Discussion with A. Vezza. 6/89. 

5 "Machine Recognition of Hand-Seni Morse Code." by B. Gold. Trans. IRE, PGIT, IT-S. 1959. p. 17. 
^ Discussion with R. Aide, 5/89. 
A. Vezza, ibid. 


only, which was what the MCR project wanted to do first, without going "on the air." As 
a consequence the "artificial ham" station was never built at MTT. 

As a result of the LCS discussions and enthusiasm a proposal to design and 
construct a computer MCR, named COMCO-1, was made to and approved by ARPA as 
part of the LCS effort in 1974.5 xhe MCR project quickly became the major effort of the 
LCS programming technology group. The rapidity of responses on both sides probably 
reflected the high level interest in ARPA and the strong desire of the IPTO group and the AI 
community to "bet on a good horse," at this timc.^*^ 

The MCR problem was categorized into general domains clearly described by die 
leader of the MCR effort A. Vezza, the leader of the projectii 

For purpose of organizing our thinking on the Morse problem, we have 
conceptually divided it into four domains over which processes must work 
and for which we must have models of expertise. One should keep in 
mind, however, that a human operator does a marvelous job of integrating 
the individual processes into a singly whole process, indicating a close 
interrelationship between the domains into which we have fragmented die 
problem. The four domains over which processes must perform and for 
which we must have a variety of models arc as follows: 

a. The Morse transcription environment -- This domain contains models 
and processes for correcUy transcribing sequences of dots, dashes and 
spaces in their symbolic representation, that is, outside the radio 
environment In order to do the task properly, processes must have a 
knowledge base of the donoain of discourse. For instance, if COMCO- 1 
is in a negotiation phase with another operator, then the processes 
transcribing the Morse must have knowledge about die protocol and 
special macro symbols used in negotiation in order to transcribe the 
signal correcdy. The structure of a message must be understood if die 
header, body, and signamre are to be properly transcribed and die word 
count checked. Similarly, die processes must at least have knowledge 
of a reasonably sized lexicon in order to properly perform the 
transcription of die body of die message. (The tacit assumption is di« 
the message is not ciphered- However, if ciphered Morse were to be 
handled, dien one would need not die lexicon but radier die length of die 
cipher groups, die group count and die characteristics of die class of 

8 This idea is discussed in the earliest 1974-75 LCS progress reports of the MCR project 

9 LCS Progress Report XH, July 74-75, MTT, p. 107, contains a general description of the problem 
prospective application of "expert system" technology to it 

10 R.Kahn,ibid. 

1 1 LCS Progress Report, ibid., p. 1 10. 


operators associated with a particular network which sent the cipher 

b The radio environment - This domain contains ^^J^P^??!^^^ 
for the radio environment Here exist models of: how mdividual Moree 
sound in terms of tone, drift, chirp, hum, etc; the effects of 
environmental conditions, such as fade, mulupath, etc.; the effects of 
interfering signals, how to deal with diem and when signals c^^^ 
cannot be separated properly into individual signals. Clearly these 
processes must provide the abiUty for receiver and transnutter mmng 
and for tracking signals. 

c The Morse network environment - This domain contains models and 
processes for understanding the special network negotiation lanpage 
used by operators in a Morse network. In this domain the models and 
knowledge must be most complete in addition to a lexicon of the 
vocabulary, understanding of the syntax and sexmnncs of the language 
is required in oider to understand the meaning of what is being said. 
The task is compUcated by the fact that not only arc most words of the 
vocabulary ambiguous, but even what one could term a clause or a 
"sentence" can ^ ambiguous. Thus, a rather global view of what is 
being said is required in order to understand what is transpinng m me 
Morse network environment 

d Sender recognition - This doniain contains models and processes for 
' recognizing a sender, if possible, and providing information about his 
or hCTidiosyncrasies. to aid the processes of transcnption, signal 
tracking and understanding. Typical kinds of mfonnation that help 
identify operators are the statistical variance of a pamcular operator s 
rate, the piocUvity for a particular operator to deviate ftom the netwOTk 
negotiation protocol in a particular manner, and the probability that a 
particular operator mis-sends 'AN* as *P*. 

The initial approach was to use MAUDE to get a first order transcription, to which 
corrections were applied such as "mark run length" - the number of dots and dashes in 
words, which had some success on sample Morse Code records. A littie later, it was 
found desirable to add a phase-lock loop signal processing system to more accurately 
determine a signal's mark and space lengths and to simulate, to some degree, the abiUty of 
a human operator to identify a specific sender's transmission. TTie output of this filter fed 
into a MAUDE decoder. A vocabulary (later, vocabularies) of EngUsh words and of die 
radio operators' standard language (Q Signs, Pro-signs, call-signs, headers) was compiled, 
and AI techniques of lattice search applied in an approach, called COMDEC, to 
systematicaUy identify alternative word translations. Further elaborations were made to 
COMDEC, applying grammatical rules, and eventually incorporating AI "augmented 
transition network" (ATN) techniques to die resulting sentence options. A somewhat 


simUar set of procedures was adopted in CATNIP, which dealt with the Q-language and 
message header structure. 

Figure 1 oudines the relation between these major modules of the COMCO-1 
system as of 1977 (some two-years into the project). About Fig. 1 the project leader 

Figure 1. The Three Major Modules of the Morse Code System and the Domain 

Models They Use 

12 LCS Progress Report, XV, July 1977-78, p. 197. 


Figure 1 shows a block diagram of the three major modules of the Morse 
Codt system COMCO-1. Also shown are the necessary domain models 
required by each module in order for it to perform its task properly. 1 ne 
vl^vy Une in the diagram indicates that the signal processmg system, whicn 
is composed of special hardware and a PDP-1 1 computer, is not miegrated 
with the other major modules which are COMDEC, the transcription (or 
translation) module, and CATNIP, the chatter and header undemanding 
module. The last two are software modules wntten m MDL. (A oar-iuce 
language) and running under TOPS-20 and ITS. Experiments are 
conducted independently for the signal processing system, and huinan 
intervention is required to transfer the results to the other wo modules. 
COMDEC and CATNIP are well integrated, with appropnate feedback, and 
externally they appear to behave as one system. 

The MDL programming language had been developed earlier by the same group, 
when working in automatic programmmg. 

Eventually CATNIP included the ATN module for COMDEC as weU as the 
"chatter" of Q-and Pro-Sign and headers, and was also able to interact with COMDEC 
regarding quality of translation and storage of results for further examination.^^ MAGE, a 
further extension of the CATNIP ATN grammar, was constructed to handle additional 
words and phrases. FinaUy, the CODEPARSE "expert" module was added to handle 
transcription of Morse Code "groups," not subject to the same strucmral analysis procedure 
as word groups. CODEPARSE used such information as the number of marks and 
spacings consistent with code groups of a uniform number of characters; the use of 
numbers or alphabetic characters, but not both, in all groups; the number of code groups in 
the message, if known; and the end of the message. Despite this small set of rules, 
CODEPARSE apparently was often more successful than human operators. 

The COMCO-1 system was tried out in numerous experiments using tapes supplied 
by various groups including the Anny and radio amateurs and the environment of an actual 
HP network was simulated early on (1975) using these tapes in a laboratory setting.^^ 

The MCR project results were briefed at DARPA in fall 1978. There had been 
earlier briefings, and considerable interaction with S. Squires, then of NSA. over a period 
of about a year. The NSA computer laboratory group was soon able to simplify the MIT 
results and reprogram them in a more precise language, more suitable for practical use.i5 

13 LOS Progress Report XVI, July 1978-79, p. 201. 

14 LCS Progress Report Xffl. July 1975-76. This was done instead of the original plan for an aruficial 

1 5 Discussion with Dr. S . Squires 5/89 and A. Vezza 6^59. 


As far as known the MCR project did not impact the commerciaUy available Morse Code 


The MCR project originated in a question raised by DARPA Director, Dr. G. 
Heilmeier, and put by D ARPA's IPTO to a group in LCS at MTT, whose capabiUties were 
intimately known. The problem was a very good fit to these capabiUties and the LCS 
group "took off. DARPA's role was to fund, approve and ensure that the results were 
communicated to NSA. The MCR project is an example of successful, efficient AI 
appUcations technology transfer to a laboratory group in an operating agency. Because of 
the competence of this laboratory group and the facilities available to them, the 
communication and assimilation of results was very efficient. Dr. Squires stated that die 
last year of MIT's work was in fact not necessary, because die NSA group had by tiien 
already replicated and improved the (primarily software) product .^"^ 

Apparentiy no "breakthrough" or new AI research was needed. A. Vezza states that 
he felt confident, after the first three months, that they could solve the problem to a 
satisfactory extent using techniques that were available. He terms it a "toy" level problem, 
compared to that of EngUsh language translation.i8 Several student contributions were at 
the Master's thesis level. 

Vezza feels that it is very unusual in his experience to have a problem that "came 
down from the top" lend itself to this type of solution and efficient transfer.^' 

MCR's success also helped the credibUity of the AI program generally. Dr. 
Heilmeier required a review of the IPTO AI program by the JASONs which Kahn 
describes as a "Uttie bit of a confrontation. "20 However, Vezza also briefed die JASON 
group and had no difficulties with thenL^i 

Vezza also credits much of the success to the fact that diis project had a single, weU 
defined objective, was earned out by a single group under a single leader and had very 

A. Vezza, ibid- 
Squires, ibid. 

18 Vezza, ibid. 

19 ibid. 

20 Kahn. ibid. 

21 Vezza, ibid. 


good communication at a technical level with a competent "user" group leader. The fact 
that the LCS group involved was intimately known to the ARPA program manager at the 
outset probably enhanced the efficiency of stan-up. which also added to the probability of 

The MCR project cost about $2 miUion and was not funded separately from the 
LCS "umbrella" xaskP- 

22 A.O. 2095 of 1/72. 









MCRJ^ mm m 





- 1 

I I 
I ■ 















7-24-89-11 M 




In mid 1976, DARPA and the Navy (NAVELEX) began a joint five-year program 
to speed up the application of new artificial inteUigence. computer, and networking 
technologies into the miUtary command and control area. The centerpiece for this program 
was the Advanced Command and Control Aichitecmral Testbed (ACCAT) facUity which 
was located at NELC (later NOSC), near their "Warfare Evaluation Simulator" in order to 
allow interaction with the war games going on there. ACCAT included prototype mobile 
remote terminals linked via satellite by a secure subnet of the Advanced Research Projects 
Agency Network (ARPANET), In addition to demonstrating and testing new technologies 
using AI techniques for distributed, relational data base naanagement and natural language 
query, ACCAT was also a testbed for extending ARPANET to some types of militarized 
computers, and of ^)proaches for ARPANET security. While specific ACCAT influence is 
hard to trace, recent renewed command control (C^) efforts with AI technology and similar 
objectives indicate its positive influence. 


In 1976 the Navy Electronics Systems Command (NAVELEX) had C? projects 
under way to develop a prototype task force command center (TFCC) and a fleet command 
center, with supporting efforts at NELC and NRL, and related research on decision aids at 
ONR. The Defense Communications Agency (DCA) had also begun efforts toward tiieir 
AUTODIN n for data communications. The Navy projects soon ran into difficulties in 
interfacing the different types of computers involved in their C^ systems. ARPANET 
technology offered a way to deal witii this problem, but had not yet been implemented on 
militarized computers such as the UYK-20. There was also an appreciation in the DoD that 
C2 had lagged in making use of applicable state-of-the-art technologies. 

Dr. G. Heilmeier, who became ARPA director in 1975, also felt strongly that the 
DARPA-supported efforts on AI technology should be directed more towards appUcations 


such as the Navy needed.^ As a result of discussions with Chief of NAVELEX. DARPA ^ 
and the Navy signed a memorandum of agreement for a five-year program beginning in FY 
1976 to set up ACCAT, a testbed at NELC incorporating the most up-to-date computer, . 
networking, and applicable AI technologies. 

Preliminary ACCAT activities consisted in obtaining DEC KA-10-2. 11-20, and 
20401 computers, TENEX and UNIX operating systems, installation of these at NELC. 
a!nd arranging ARPANET interfaces with necessary security. There were a number of- 
challenges involved, including setting up ARPANET, which had "grown up" on . 
commercial computers, on militarized computers such as the UYK-20, and providing 
security systems with ARPANET bandwidths.^ Prototype "mobile" terminals to be ^ . 
linked witii ACCAT in a sateUite with ARPANET were set up at tiic U.S., Navy - 
Postgraduate School and the Fleet Numerical Weapons Center at Monterey. This AGCAT-^, ^ 
effort on networking techniques appears to have had some impact on tiic Worldwide y 
MiUtary Command and Control System (WWMCCS), which was dealing with similar 
problems at that tinie. 

The University of Southern California's Information Sciences Instimte aSI) was:" ^ 
linked to ACCAT by ARPANET to provide additional computer support and otiier. : - 
services, A study was also undertaken by NELC to define prospective tasks for' ACCAT.3 . 

One of the first ACCAT tasks in 1977 was a typical C^ problem of obtaining timely?: 
information from distributed data bases at the Fleet Command Centers in Hawaii and v 
Norfolk. The ACCAT approach to this problem involved appUcation of new rclationaf arid;; 
distributed data base management and query technologies. A modification was made of 
Computer Corporation of America's SDD-1 system for management of relational, 
distributed data bases. The extensive data bases were to be handled by "modules" of the?^ i 
Datacomputer, developed by the same company with DARPA support, initiaUy to provide a 
'very large storage memory for seismic data developed by the programs of DARPA's 
Nuclear Monitoring Research Office, The data modules were linked via sateUite and:^ 

1 Testimony of Dr. G. Hcilmcicr, Hearings before the Committee on Anncd Serjices. H,O.R.. for 
Department of Defense Authorization and Apfwopriations for FY 2976 and 1977. 94th Congresi, 1st 
sessions, Part 4, p. 4908. 

2 DARPA archives for AO 3175 of 1/76 and discussion with D. Small 5/88. 

3 "A Digest of Research Applications for the Advanced Command and Control Architectiual Tcsibed 
(ACCAT)." by D.C. McColl, NELCTN 3198, 1976. 


ARPANET. Some fleet data bases were simplified and '•sanitized" for use over the 
unclassified sections of ARPANET. To assist personnel not famiUar with the data bases in 
infonnation searching, the Rand Intelligent Terminal Agent (RITA) AI program for file 
search, previously developed by the RAND Corporation with DARPA suppon, was 
implemented on the ACCAT computers.^ To deal with the further problem of access to the 
data bases by personnel unfamUiar with computers, a new "naval vocabulary" was 
incorporated in the LADDER natural language interaction system also set up at ACCAT by 
SRL The new ACCAT capabiUties also involved advanced display systems which were to 
be used in connection with simulated war games played on the Warfare Evaluation 
Simulator at NOSC, the reorganized NELC These displays could aUow interaction and 
evaluation by both fleet and laboratory personnel^ 

The results of working with ACCAT generally indicated the potential of the new AI 
technology.^ But limitations in a number of the technologies involved soon became 
apparent. For example, whUe the SDD-I modification would allow some ACCAT data 
base management, its speed was limited because the ARPANET coimnunication bandwidth 
limited the rates of exchange of data between data modules. Also, problems of consistency 
and concurrency of the relational data base management system were not completely 
solved. Eventually, only one large data base, on one Datacomputer, was used by 

This ACCAT experience with relational data bases appears to have been one of the 
earnest. It appears to have had some impact on later work by Computer Corporation of 
America (CCA) which led eventually to the M-204 relational data base management system, 
now implemented on IBM 9370 computers and used in several military appUcations 
involving localized, but not distributed, data bases.^ 

The CCA SDD-I experience also seems to have had some influence on standards 
for data base management systems and also on a current effort (written in Ada) for Army 

4 Discussion with D. Small, NOSC, 5/88. 

5 D. Small, ibid. 

6 R. Bradenburg. NOSC, discussion 5/89. 
R. Brandenburg, ibid. 

* D. Small, ibid. 


data base management The ACCAT experiments can be credited with showing Navy's 
C2 systems builders how to use relational data bases.^ 

A localized relational data base with a corresponding display is now used in the data 
base management systems in the Navy's Developmental Task Force Command Center, and 
in the ship's data management system (SDMS) tcstbeds on the earner U.S,S. Carl Vinson, 
supported by D ARPA and ONR. 

Some of the other technologies used in ACCAT had less success. The RTTA 
system was implemented in ACCAT, but after some early trials seems to have had little 
use. One of the early trials, on a simple navigational problem, indicated RTTA was slower 
than the standard manual procedure. The Language Access to Distributed Data with Error 
Recovery (LADDER) natural language system, was also used together with SDD-1. 
However, after a few trials the conclusion was drawn that its capabilities were too 
Umitcd.10 The current prototype Tacdcal Flag Command Center (TFCC) at NOSC does not 
use a natural language system. The strategic computing program for a facility at 
ONCPACFLT, however, now includes a new natural language system. 

One of the main recommendations from the NOSC planning smdy was to exercise 
ACCAT in a large experiment using Planning Research Corporation's SURVAV Decision 
Aids programs to simulate ships' routing to minimize detection by sateUite." Ttas exercise 
was run, but SURVAV does not seem to have been used subsequently in war games. 
However, ACCAT terminals and faciUtics were used in NOSC war games during the 1978- 
198 1 time period. ACCAT computers and the ARPANET connections made avaUable by 
the project were also capitalized on extensively by NOSC for its own projects and are still 
used today. 

DARPA participation in the ACCAT joint project terminated in 1981 and the 
ACCAT facility was transferred to the Navy. For some three years thereafter, apparentiy, 
Navy funding was not available, and the ACCAT facility was not used. In the period 
1984-1987 a copy of the USS. Carl Vinson's data base management system was installed 
in the ACCAT space. Near the end of this period, the ACCAT facility was replaced by a 

9 ibid. 

10 ibid. 

11 A.0. 3958. and 4430. 


new C2 testbed incorporating more recent AI techniques, but in a conservative fashion, and 
using extensive local area networks.^^ 


ACCAT apparently originated in high level discussions between the DARPA 
director. Dr. Heilmeier, and Navy officials anxious to make more rapid progress in C^.^^ 
It was not an Information Processing Technology Office (IPTO) initiative, R, Kahn states 
that while Dr. Heilmeier pressed hard, there was no way to get him what he wanted at the 

CDR F. HoUister came from the Naval Electronic Systems Command (NAVELEX) 
to run the project It is not clear, however, that mid-level NAVELEX support was 
enthusiastic. There were multiple objectives: to test current AI and related technologies, 
acquaint those in C^ R&D with their potential, and to challenge AI researchers to come up 
with useful applications. ACCAT, which formerly transferred to NOSC, did not, 
apparently, lead directly to adoption by the Navy of any of the AI technologies specifically 
implemented or even to immediate foUow-on projects. It did allow some degree of test of 
those technologies attempted to be applied and in so doing achieved many of its basic 
objectives. ACCAT apparendy stimulated a general interest at NOSC. 

The networking technology aspects of ACCAT apparendy were transferred 
effectively to the NOSC environment ACCAT also was useful for demonstrating how 
different militarized computers could "communicate" with each other and to develop 
approaches to ARPANET security. This part of the ACCAT effon apparendy was rapidly 
assimilated into NOSC. It appears also to have had some impact on the directions taken by 
the DCA's WWMCCS system with similar problems. 

Despite the lack of specific AI systems impact, recent Navy 0- programs at NOSC 
are trying again to incorporate some AI expert systems. This new program seems more 
conservative and uses a less ambitious data base management systems than ACCAT. The 
DARPA Strategic Computing joint project with CINCPACFLT, staned in 1984, also 

1 2 Discussion with LCDR Ted Krai, 7/89. 

R. Kahn, p. 247 in Expert Systems and Artificial Intelligence^ Ed. T. Bartee. Howard Sams & Co. 

Kahn, ibid. 

1^ CDR F. Hollister, "ACCAT: A Tesibed for Exploring Change," in Journal of Defense Research, 
Vol. 78-1, Jan. 1978, p. 39. 


appears to have many of the same kind of objectives as ACCAT. for its complex of AI and 
computing technology. 

The lack of Navy momentum in the early I980's is attributed by some as a 
consequence of the small degree of involvement of fleet personnel. It is difficult to get fleet 
people seriously involved when away from operations.'* Partly, it may have been due 
also to skepticism by mid-level NAVELEX staff. The performance capabilities of the then 
available AI technology was very much stressed by the ACCAT. Whether this challenge 
inspired new advances in AI technologies is not clear. Some key Navy personnel feel that 
there are problems with a testbed approach to C\ and do not expect any kind of "quantum 
jump" in performance. Their view is that improvements in C^ should be cautiously 
evaluated and developments expected to be more "evolutionary."" Perhaps for reasons 
such as just mentioned, DARPA-Navy CINCPACFLT testbed experiments are run m 
parallel with the regularly operating systems, by fleet personnel.'* The testbed gradually 
has been taking over some of the operational load. 

From project records. DARPA's outlay for ACCAT was $15.7 million. 
NAVELEX outlay, for the five-yean to 1981, was about $1.5 miUion. 

1< Discussion with CAPT R. Manin, 7/89. 

1 ^ Discussion wilh R. Le Fande, Office of the ASN R&E. 5/89. 

18 R. Manin, ibid. 














H > 









*2f*;i - " " ? i' 

r. . 1 ■ 1 



• • • • 
• I 





•r • 
























7-24-89-1 OM 






acoustic frequencies. Such arrays were used in the Navy's Interim Towed Array 
Surveillance System (ITASS), which was operational in the late 1960s 2 

Auiand's proposed objective to explore the coherence of acoustic signals over wide 
apertures, together with the favorable propagation expected at low frequencies, had been a 
matter of discussion by those active in the area for some time, and dovetailed with new 
ARPA interest in exploring the limits of submarine detection systems in the ocean.3 The 
fact that much of the technology for this phase of exploration was nearly off the shelf and 
might be low cost, were additional incentives. There was some technical risk, since 
previous measurements by BeU Laboratories indicated that usable apertures might be 

ARPA responded quickly with funds to rent a modified seismic towed array 
(together with the handling and towing gear) and the towing ship itself.^ This ARPA- 
sponsored activity excited some Navy interest, and the Navy's NAVELEX ASW 
surveillance office (PME 124) provided funding for modification of the on-board analog 
processing equipment. ARPA further prescribed that sophisticated digital processing 
methods be also applied off-line.* 

The first at-sea experiment in a low noise environment with the long seismic array, 
rented fixrai a commercial geophysical exploration company, gave spectacular results. This 
success quickly led to the establishment of a joint R&D program and a formal steering 
committee for the pioject, with equal funding from the Navy and ARPA. The technical 
problem for this steering committee was to choose between extending the length of 
telemetry-type arrays then being developed by the Navy, for the SURTASS program, 
versus towing the seismic arrays at greater depth than had been used in their geophysics 
work. The shorter Navy arrays had been towed at desirable depths, and had been refined 

2 Discussion with G. Boyer, Engineering Research Associates, May 19^^ 

3 ARPA had recendy been assigned a responsibility for a research program in Fleet BalUsuc Missile 
(FBM) Submarine vulnerability, by DoD. An ARPA contractor studying opuons fw" tje new pro^ 
attended one of Aurand's presentations to the Navy, in summer 1971, and recommended that Auiand go 
to ARPA with his proposal. Discussion with R Aurand. NOSC, April 1988. 

4 Discussion with H. Auiand, April 1988. Aurand felt initially that the LAMBDA arrays might in fact 
be too long, but they would find out how much aperture was useful by expcnment 

5 ARPA Older # 2001, "LAMBDA." of 12/2/71. for SIOOK. 

6 In the mid to late 1960's. ARPA had funded development of such processing techniques f 
underground nuclear tests. Application of the geophysical procesang techmques f » 

results, however, did not prove useful. Discussion with H. Aurand and T. Ball at NOSC. 4/7/88. 




The ARPA Large Aperture Marine Basic Data Array (LAMBDA) program used 
available geophysical seismic airay technology to demonstrate the potential of large acoiistie 
apertures for ocean acoustic surveillance. The first LAMBDA results decisively influencedi^ 
the Navy to lengthen the towed arrays developed for iis^-Surface Ship Towed Array" 
Surveillance System (SURTASS). LAMBDA'S performance and technology allowed the 
Navy, in 1978, to make a timely switch to the seismic technology to complete its 
evaluations and obtain DoD approval for SURTASS. 

1 Hi 



The Navy had developed towed arrays (strings of acoustic transducer-rpceiversv 
connected to processors on board the towing ship) for submarines^beginning with an ONR 
program in the early 1960's, and a Uttie later for surface ship, short-range tactical ASW. In 
the laie 1960's the Navy was beginning a program to develop arrays to be towed by surface, 
ships for longer range submarine surveillance, using technology which was an extension of 
that used in the earlier Navy systems. ; t 

Based on some preliminary ocean acoustic noise measurerfiehts using a Ibnlg^^^* 
moored, laboratory-built array, together with information on long towed arraytof th^^^ ^ 
used for science exploration by oil companies in the early 1960's, a propos^ m t^adi^ ' 
: ARPA by H. Aurand of Naval Ocean Systems Center (NOSC). Tlie proposai^ras to 
obtain and modify such a long seismic array for deeper tow than the practice in seismic 
exploration surveys, with associated low frequency signal processing, for measurin^^^ 
coherence of long-range acoustic propagation and noisc.^ Previous attempts by Aurand tQ , 
obtain support from the Navy for his proposal had not been succesM'. Apparcntiy, jh^ 
Navy's NAVELEX was mainly interested in shorter towed arrays, for use at higher . 


Aurand had previously worked on the Office of Naval Reseaichproject SEA SHD^; a moored ^ 
an^to meas^ acoustic coherence at favoiaWe ocean depths. This project failed, due to deep moonng 


to have low noise characteristics. The noise properties of the seismic arrays, when towed 
at depth and at acceptable speeds, were then unknown. 

The initial approach of the joint program was to extend the telemetry array 
technology then under development to longer dimensions.'^ This, however, soon led to 
difficulties, and as a result a new seismic array, the first LAMBDA, was built with DARPA 

The LAMBDA technology incorporated the same array structure, strengthening 
members, skin materials, and hardwire connectors as did the geophysical seismic 
exploration arrays, and was built by the geophysical exploration service companies in the 
same shop as were their seismic arrays. There were some differences: in transducer 
"loading," and in the anangements for deeper towing than for the geophysical arrays. The 
depressor for the deeper tow had been developed earHer, in 1968, by Aurand, then at 
Lockheed, for an ONR research program. There were also differences in economics, due 
to the fact that commercial competition had led the geophysics industry to low-cost, robust 
systems. Compared to the telemetry arrays, however, the hardwired seismic arrays had 
larger diameters, were heavier and had a limited number of channels for data transmission. 

The joint program entailed a combination of ocean-acoustic measurements, the 
Long-Range Acoustic Propagation Program (LRAPP) under ONR. together with 
engineering tests and exploration of operational utility of the towed arrays. In time, the 
latter two motifs dominated the more fundamental question of limits of useful apcmire.8 
The LRAPP program, however, indicated the practicaUty and robust quality of the 
LAMBDA technology. 

During this period, the Navy's SURTASS program continued efforts to extend to 
longer array lengths the approach derived from the telemetry array technology which had 
been successfully used in shorter towed arrays. Full-scale development for SURTASS 
was approved in 1974. However, difficulties were encountered with the telemetry array 

The Navy had used hardwire technology, as well as telemetry ^^.^^ 

array work. The telemetry approach had won out in a compeuuon for a total sy«em. including data 

processing, etc.. in addition to the towed array. Communicauon from H. Cox. i W 

Aurand. however, left the program because he felt it was not sufficienUy oriented toward research on 

limits of coherence in the ocean, as he had onginally proposed. 


that was being tested and in 1978 a major failure occuired.«> The SURTASS program, then 
managed by Capt. H, Cox, who had previously been in charge of the DARPA program 
also had a number of serious software problems, besides that of the telemetry The 
availabiUty of a LAMBDA type anay, and the confidence in its performance, led to a quick 
adoption of this technology for the remainder of the SURTASS program evaluation. The 
LRAPP experience, together with the positive results from the evaluation of the SURTASS 
LAMBDA-type array, were also helpful in obtaining DoD quick approval for production of 
SURTASS in 1981, without a requirement for a new array R&D program as normally 
would be the case for a major shift in technology. Such a R&D program would have 
caused considerable further delays.^! 

LAMBDA 1, the original LAMBDA anay, was given to the AustraUan government 
under a cooperative program for ASW research. In all, three LAMBDA arrays were built 
and used in the LRAPP program. LRAPP continued until the late 1970's. ONR continues 
long-range acoustic propagation research in the Advanced Surveillance Experiments at Sea, 
(ASEAS) progranL 

In 1974. DARPA set up its SEAGUARD program, a large-scale effort to explore 
tiie limitations placed on ASW surveillance tiiat result from ocean stnicture and dynamics. 
SEAGUARD involved theoretical woric, construction of a very large fixed array, ocean 
measurement and array technology (OMAT), and experiments linking fixed and LRAPP 
mobile arrays (the fixed mobile experiment [FME]), with the nXIAC IV signal processing 
capabiUtics at die Acoustic Research Center (ARC) at Moffett Field. While OMAT gave 
some valuable data, the ocean engineering problems concerning die stable deployment of a 
very large undersea array, togetiier with appreciation of the vulnerabiHty of such a large- 
fixed system, cvennially led to its discontinuance.^^ The ILLIAC IV was very effective 
when operating, but reliable real time processing was not possible, owing to its many 
breakdowns. 13 The FME, after delays, was successfully concluded by die ARC, 
however, using several PDP-10*s run in parallel 

9 Hearings. Subcommittee on DoD AppropriaUons. H.OJI. 96th Congress. 1st Session. Part 6. 
p. 1147. 

10 These problems were overcome in a straightforward program under Capt. Cox. Cf. HOR Heanngs. 
ibid, 1/6Z 

1 1 Senate Amed Service Committee, Hearings, FY 79, pL 6, p. 2998. 

1 2 Discussion with R. Cook, and Capt H. Cox, ibid. 

13 Discussion with E. Smith. ex-ARC Director. 7/88 and R Auiand. 4/88. See Chapter 18 on ttUAC. 


technology for the remainder of the SURTASS evaluations and for the first operational 
arrays. The software adjustments which had to be made in this switch were accepted as 
part of a broader software "fix" effort These performance factors were also important in 
getting DoD approval in 1981 for SURTASS production, without the normally required 
new R«S:D program to develop and test a new array. The additional ARPA funding of 
~ $12 million was needed (together with a comparable Navy outlay) in this period to 
develop this seismic array performance information. 

LAMBDA was not a hi-tech program. In fact, the Navy's telemetry array approach 
involved riskier technology. This telemetry array technology has become more robust, and 
is now used in the newer SURTASS telemetry arrays. The LAMBDA seismic technology 
was good enough to save the SURTASS program at a critical juncture. 

Aurand's motif was to get a low-cost, low-risk tool for addressing the fundamental 
question of maximum useful aperture in the ocean. However, Aurand's original plan to 
conduct a program of ocean measurements using LAMBDA, was apparently only panly 
carried out in LRAPP-ihe priorities of engineering and operational experiments won out 
OMAT, a fixed system, was not altogether successful in answering this important question. 
ARPA's FME also provided some important information on coherence of acoustic signals 
between widely separated points. Recently, however, due to the Soviet submarine quieting 
threat, Aurand's original LAMBDA (and OMAT) questions about maximum useful 
apertures have arisen again, and are being addressed in new programs. 

The DARPA ouUay of $12 million for LAMBDA does not include die later funding 
for MFA. the FME, or OMAT. 

Estimated life cycle costs for SURTASS, including the special T-AGOS ships, 
were about $2B in 1980.16 

1 HASC Hearings, ibid, p. 1 13 1. 


In the 1970's, DARPA played a major role in developing the Medium Frequency 
Array (MFA). MFA was a modification of the LAMBDA-type array and associated 
processing which extended the frequency range of the array to improve signal-to-noisc 
characteristics. 14 The MFA has been transferred to the Navy and has been used in several 
Navy R&D projects. The MFA technology also had some impact on the design of the 
improved SURTASS scheduled for deployment in 1988.15 


The LAMBDA concept and some pertinent preliminary data were brought to ARPA 
by H. Aurand of NOSC. This was very timely because of a new DoD assignment to 
DARPA on SSBN vuhierabUity. Aurand was "found" by an ARPA contractor who was 
engaged in a smdy to scope approaches to the new DARPA program. Aurand's suggestion 
that existing low risk seismic array technology would provide a way to explore the utiUty of 
large aperture acoustic systems got a quick response from ARPA. This "seed" money 
probably would not have been obtained from die Navy for some rime, since the Navy did 
not respond positively to Auiand's proposal. The first $100,000 ARPA investment clearly 
showed that the use of long arrays to conduct surveillance at low frequencies was 
promising, and might be achieved at lower cost than many had believed possible. The ' 
Navy reacted quickly to participate in a joint exploratory program and to revise its plans for 
SURTASS toward longer arrays. This decisive step toward longer arrays was probably the 
major impact of LAMBDA. 

However, the Navy did not then adopt the seismic technology for those longer 
arrays but continued along the direction it had been going in SURTASS with telemetry 
array technology. There were trade-offs, and the Navy apparenUy fch that their experience 
with the deeper telemetry arrays and the apparent advantages of such arrays outweighed the 
difficulties the joint program had experienced carUer witii the first long telemetry array. 
EvenmaUy. after the SURTASS telemetry array faUed at a critical stage of its evaluation, the 
Navy mmed, in 1978, to the seismic array technology. The facts that the then SURTASS 
program manager, Capt. H. Cox had previously been in DARPA, and was thoroughly 
familiar with the performance of die seismic technology in LRAPP and other tests, together 
with the availability of an array for test, were key factors in switching to the seismic array 

14 AO 3447 of 6/77. 

15 Discussion with Capt H, Cox, 6/88. 


Figure 1. SURTASS 




















N 19 








AO 2001 




1 1 













7-31 -69-1 1M 




Building on earUcr Navy and DARPA efforts, in 1978 a joint D ARPA-Navy project 
began with the objective of achieving a laser communications Unk between aircraft, space 
platforms or mirrors, and submerged submarines. The ground-based laser-space minor 
pan of this effort built largely on efforts toward high powered visible lasers in the DARPA 
Suategic Technology program, and developed techniques for compensation of amiospheric 
propagation effects which were transferred to the SDIO. An efficient laser-receiver and a 
narrowband, matched-wavelength excimer-Raman converter laser system were developed 
and used in successful demonstrations of aircraft-to-submerged-submarine com- 
munication, in 1988, after transfer of the Submarine Laser Conmiunications-SatelUtc 
(SLCSAT) program to the Navy in 1987. 


The existence of a favorable wavelength range in the blue-green for optical 
transmission in the sea has been known for a very long time. The potential of a suitable 
laser in this spectral range for conmiunicating with and detecting submarines was 
recognized soon after the discovery of die laser in die early 1960's. However, for some 
time it has proved difficult to find a practicaUy useful laser in this wavelength region.^ In 
the early 1970's Navy Electronics Laboratory Center (NELQ, later Navy Oceans Systems 
Center (NOSC) commenced an effort, with ARPA support,^ to develop an optical system 
for communicating between aircraft and submarines, using available high power arc lamp 
sources. This led, in the 1971-75 time period, to NELC's Submarine Air Optical 
Communication System programs in the 1971-1975 period, which also included 
exploration of two-way communications between aircraft and submarines. Results of this 

1 One of the earUest lasers, found in 1961 by Gould at TRG under ARPA sponsorship, was the green 
copper vapor laser. WhUe further developmcni to reduce power demands has led to its use for a major 
approach to laser isotope separation and other commercial uses, it has not yet proved practical for Navy 
communications use. 

2 A.O. 1871. 


early work underlined the need for more powerful and efficient blue-green light sources 
and sensitive receivers. 

In the late 1960's, the Lincoln Laboratory had developed atomic vapor resonance 
receivers for optical communications systems, and had recognized the potential of the 
Cesium vapor as an atomic-resonance filter (ARF) receiver in the blue-green for the Navy. 
Proposals to carry out fiarther development were made by Lincoln Uboratory to the Navy 
and others, but no interest was found and die Lincoln group nimed to other things.^ 

In the mid 1970's the Office of Naval Research (ONR) and NELC began Optical 
SateUitc Communications (OPS ATCOM). aimed at eventual use of lasers in satellites for 
communicating with submarines. In this project the sun was used as a source to inakc 
measurements of the characteristics of Ught penetrating to increasing depths in the ocean. 
In 1977, a study was made of the relevant state of Uie an of electrooptical devices and 
associated light propagation modeUng.^ The resulting OSCAR program was mainly - 
concerned with lasers in aircraft to communicate with submarines, since the high powers 
required and corresponding state-of-the-art sizes of the blue-green lasers seemed to rule out • 
space systems. However, as part of OSCAR long range smdies were made by industry of 
ground-based lasers and space mirrors, and space-based lasers for future systems. The 
potential utiUty of an atomic resonance narrowband filter optical receiver was mentioned, 
but not emphasized, in the 1977 report.^ 

In 1976, the advantages for a laser receiver of properties of a Cesium vapor atomic 
resonance fUter (ARF), with narrowband sensitivity to blue Ught and a fluorescence in the 
red, were rediscovered by Marling at the Livermore Uboratory, in an effort suggested by 
the Navy.« Excimer lasers, having emission in the ultraviolet, began to be investigated in 
the early 1970's, initially using powerful large c-beam exciters but with generally low 
efficiency. In the late 1970*s. a more compact discharge mode of excitation was 

Discussion with R. Lemer, Lincoln Laboratory 9/88. 

Technical Chionology of SateUite Laser Communications (SLQ and Related Efforts, ORI Technical 

Report 259, 9 March 1987. 

In 1978 a McDonneU Douglas study of Cs atomic resonance receivers was conducted which stated that 
no matching wavelength space-qualified laser was available. 

Testimony of Lowell Wood. LLNL, to the R&D Subcommittee of the Senate Armed Semccs 
Committw 5 At>r. 1979, p. 3326. Wood describes the origin of the LLNL involvement m the 
SS^^iomX^^^^^ problems as due to a challenge by S. Kaip of the Nav^ ^^-^^^^^ 
Center (NOSC) to develop a suitable receiver. Wood also oudmes a ground-based la?e'/submanne 
communication system concept and suggested program plan for a GBL system exploiimg the LLNL 
ARF development 


demonstrated at the Naval Research Laboratory (NRL), which had been working on 
excimer lasers with DARPA suppon. NRL also found a way to increase efficiency by 
adding HCl as a CI supplier for the XeCl excimer haUde laser. A little later, conversion of 
the XeCl transition into the blue by "Raman" conversion in an oscillator-cell involving lead 
vapor was discovered at NRL and a little later at Northrop.'' 

In 1977. ONR opened discussions with DARPA to form a joint Navy-DARPA 
project to investigate laser communications to submerged submarines.^ Earlier, the Navy 
had developed an Extremely Low Rrcqucncy (ELF) electromagnetic system to communicate 
with submerged submarines, but in the 1970*s was having difficulty finding an acceptable 
place to locate it. Congress was becoming increasingly sensitive to environmental 
considerations which many people associated with the ELF system, and was urging the 
Navy and DoD to generate some alternative. However, the ELF approach was relatively 
mature and the Navy had spent a great deal of time and high level effon to have it 
approved.' At this time DARPA had several ongoing programs to develop blue-green 
lasers. The largest of these was for directed energy weapons (DEW) applications in space, 
or from ground to space, and there were other efforts related to submarine detection from 
aircraft (ODACS). and for deep-sea search (DEEP LOOK). One of the main objectives of a 
joint DARPA-Navy program was to exploit these other technological developments, the 
largest of which was in the DEW area, for the communications objective. Another was to 
be able to use investigations of die lower power conmiunications laser to explore 
technologies tiiat were also of interest to the directed weapons application area, without 
having all the technical and economic problems of high energy laser systems. 

Initially, the joint program followed two approaches. One envisaged high-powered 
ground-based lasers (GBL) at locations where cloud-free upward propagation would 
occur, and mirror-satellites to reflect the laser beam down to chosen areas of the sea. This 
approach built on the previous DARPA DEW efforts toward high-power, short-wavelength 
lasers and precision, Ughtweight space optics, and on techniques to compensate for 
propagation effects due to atmospheric irregularities. In the joint program, the GBL 
approach was to be emphasized by DARPA. The other approach, emphasized in the Navy 
part of the program, involved a laser in a space platform or aircraft In this approach it was 

' Discussion with J. McMahon, NRL, 3/89. 
^ Discussions wiih D. Lewis, 4/88. 

9 "The ELF CommunicaUon System Arrives at Last," by CapL Ronald Koonu, Signai Jan. 1, 1986, 
p. 21, 


considered that a message could be sent from the ground to the elevated platform by 
conventional electromagnetic transmissions, and then sent optically from the platform to 
selected areas of the sea surface. In both approaches it was soon recognized that to send a 
message by laser pulse modulation simultaneously to very large areas of the sea would not 
likdy be practical, because of the very high laser energy and large optical systems required. 
Instead. smaUcr "spots'* on the ocean surface would be iUuminated by die laser beams, 
sequentially in time, in some random pattern covering die submarine operating area.i*> 
Common to both approaches was the need for a suitable optical receiver to be carried by the 
submarine which could selectively match, as closely as possible, laser wavelength and 
narrow optical bandwidth in order to provide more pulse signal photons dian would come 
from fluctuations of sunlight in die day or biolumincsencc at night Common also were 
questions relating to laser light propagation, including time-spreading of pulses, dirough 
aonospheric clouds and through the sea water. 

This joint program took place in several phases. The first phase occurred between 
1978 and 1982, and featured several demonstration-experiments, together widi a broad 
program investigating laser sources including frequency-doubled Nd-Yag. atmospheric and 
ocean optics measurements, and systems smdies. The first of diese experiments, in 1979. 
involved measurements of laser light transnussion dirough clouds. Some of diese 
experiments included participation by an aircraft from the Air Force Space 
Communications Project-405B, m order to determine how low dieir system, designed for 
space links, could reach in the atmosphere." Comparison of die 1979 experimental data 
widi simplified computer models of dirough-cloud transmission apparendy showed only 
fair agreement. 

In die late 1970's the University of Arizona Optical Science Centeri^ began work to 
exploit some of dieir optical coating techniques in die construction of a more efficient ARF, 

1 0 "Submarine Laser Communication," by Cdr. Ralph Chatham, EUctromc Defense, March 87. p. 63. 

1 1 EHscussion with Monte Ross. 7/88, Ref. 3, p. 2-4. 

12 "Temporal and Angular Spreading of Blue-grecn Pulses in Clouds." G.C. Mooiadian and M. Geiler, 
Applied Optics, Vol. 21, # 9. 1 May 1982. 

1 3 u of Arizona Optical Science Center was started with ARP A assistance, in the early 1960's. ^ In the 
later I960's the Air Force gave support to assure its survival. "The Optical Science Center. U. ot 
Arizona, undated. 


building on the previous work by Wood's group at Uvennore.i^ Apparently, this effort 
began as a result of a suggestion by the Navy progiam managers. 

The ARF receiver that resulted incorporated die special coatings previously 
developed by the University of Arizona, one of which (on the "top") accepts die blue laser 
light excidng the Cs, and containing the subsequent red fluorescence, and anodicr coating 
on the "bottom" contains the blue light and aUows the red to pass through to photo 
detectors. The cell contains a rare gas buffer, together with the Cs vapor, found necessary 
to adjust the partial pressure of Cs and the red line broadening to allow the optical depths in 
the blue and red lines to have desirable properties, as well as to avoid nonunifoimities in Cs 
vapor concentration due to uneven temperature distribution. 

In 1980, a memorandum of agreement regarding a program to develop laser 
communications widi submarines was signed by DARPA and the Navy. Another 
demonstration experiment, in 1981, was done by NOSC again using a frequency doubled 
l-watt Nd-Yag laser in an aircraft, this time with a receiver employing a birefringent "Lyot" 
filter and a photomultiplier mbe, mounted on the R&D submersible DOLPHIN. The wider 
acceptance angle of this filter allowed more photons to be capmred Uian the standard 
multilayer interference filter which had a narrow angular field of view, proportional to the 
filter band-pass.i6 The technical objective of this task was to obtain performance data with 
which to compare calculated results from models, using measured optical properties also 
obtained under the program. This time there was encouraging agreement between models 
and data. 

After this successful demonstration of communication from an aircraft to the 
experimental submarine DOLPHIN, NOSC studied die application of die available 
technology to communications from aircraft witii SSN's in direct support of battie group 

Also, an intensified examination was made of a number of odier candidate laser 
systems with optical output in the blue-green, such as HgBr. Toward die end of dus first 
phase in 198 1, attention began to be focused on die potential of die XeQ-lead vapor Raman 

1"* A.O. 3623 5/78. See also Fn. 18 below. 

15 The University of Arizona's new coatings were "in search for a problem" for application. The ONR 
and NOSC managers suggested the ARF. Discussion with Dr. M. White, ONR, 8/88. 

16 See, e.g.. "Detecting High Aluiude Explosions by Observation of Air nuorescence," by T.M 
Donahue, Proc IEEE, Vol. 53. No. 12. 1965. p. 2072, where problems of discnminauon against 
sunlight are discussed. 


laser system, with emissions that provided a very close match in wavelength and 
bandwidth to the blue resonance of the Cesium vapor atomic resonance filter (ARF). In 
1981-82, several industries developed competing space-based system concepts. At this 
time the program began to to be called "Strategic Laser Communications" (SLQ. 

In the second phase, roughly 1981-1983, there was greater confidence, since the 
XeQ laser efficiency was now a few percent, and the lead vapor Raman converter, in an 
osciUator-amplifier configuration, operated at about 50% efficiency. More emphasis was 
now given to improving the receiver properties. 

During the period of these two phases there were also several developments more 
specifically appUcable to the GBL approach. Thus the EMRLD laser, a state-of-the-an 
high-power excimer laser, was built primarily for DEW appUcanons, but could be adapted 
also for the GBL communications role. Lincoln Laboratory also conducted experiments at 
the ARPA Maui Optical Station (AMOS) on atmospheric transmission compensation 
techniques, which would be needed for bodi DEW and GBL applications. 

Several studies of botii types of system designs, GBL and SLCS AT, were made in 
this same time frame. Statements were made, in DARPA testimony to Congress, that a 
decision would be made in about 1984-85, as to which of the two approaches, ground^ or 
air-based (or space), would be chosen. 

Anotiier airborne-laser field experiment (SLCAIR 1984), was conducted in 1984, 
using a more powerful, high-pulse-rate Nd- Yag laser, and two types of bireftingent Lyot- 
filteis. A second MOA was also signed between DARPA and the Navy. 

When the SDI program began at this time the GBL laser technology was transferred 
to it, along with a major portion of the DARPA high-energy laser effort. SDI proceeded to 
conduct further tests of some of the GBL amaospheric compensation techniques using 
rockets, the Space Shuttic, and the (now Air Force) AMOS facility. 

From this time the DARPA program focused primarily on a satellite-bomc laser 
communications system, potentially useful in communicating, oceanwide, with all types of 
submarines.17 The next phase can be considered to have begun with the transfer of the 
ground-based part of the program to SDI and plans with the Navy for another experiment, 
SLCAIR, in 1986, to determine capabilities of communicating with a submerged submarine 

I'' AO's 3623, 4011 and 5069. An addiLional motive for choosing the space-based system 
persuasive approach to Congressional staff by a contractor interested in the space system. 


under environmental conditions that could be considered both unfavorable and potentially 
operationally important. This experiment used the same Nd-Yag green laser source as in 
1984, with two types of Lyot filter, one involving CdS with a wider field of view.i^ This 
experiment also involved "scanning" of the laser beam simulating the pattern on the sea 
surface that might occur in an actual, air- or space-based system.^^ With scanning, it was 
possible to better determine actual communications rates 20 The new program name 
"Satellite Laser Communications" began to be used about 1985. The program now focused 
chiefly on technology for receivers of high overall efficiency, including photosensitive 
materials with higher quanmm efficiencies for detection of the red Cs fluorescence, 
building on previous work by the Army's Night Vision Laboratory (NVL).2i Efforts with 
industry toward an engineering model XeQ-Raman laser-converter system, suitable for use 
in space, also intensified. 

The improvements of receiver parameters reduced the space laser output power 
required, thereby allowing the use of solar ceUs for prime space power. DARPA funded 
construction of a XeCl-Pb Raman Laser System by Northrop which had a compact design 
for space qualification. This design, however, did not permit easy access to the laser. 
Because of this it was difficult to operate the laser as designed, and tests were not 
completed by the time the Navy took over prinaary responsibility.^^ Laboratory tests of 
another (not space quaUfied) system indicated a "lifetime" exceeding 10* pulses, with a 
goal of 109. A field test in July 1988 included an XeQ Raman (but not the space qualified) 
unit in an aircraft, and a prototype ARF receiver on an SSN, and was, apparentiy, quite 

The SLCAIR and SLCSAT programs also included some effort on alternative 
lasers, notably solid state lasers that could be efficiently pumped by semiconductor diodes. 
A compact, diode-pumped glass laser constructed under this program apparentiy has been 
of considerable interest to the SDI effort Solid-state lasers of this type are considered by 

1 8 Work on CdS was apparenUy dropped because of the difficulty in obtaining sufficient material of the 

requisite quality. NOSC memo to authors, 1 1/89. 
^9 Discussion with G. Mooradian, 7/88. 

20 In Congressional testimony the average rates expected for a SLCSAT system were stated by to be 
roughly comparable to those of the ELF system. Cf. Department of Defense Appropnauon for 1984, 
98th Congress. 1st Session, Pan 8, USGPO, 1983. p. 399. 

2 1 SLCSAT requirements involve integrating photons over the receiver bottom surface area, less stringent 
than for NVL imaging devices. However, along with this improved photon sensitivity there is an 
increase of internal noise. 

22 Discussion with Cdr. R. Chatham. 8/88. and NOSC Memo. ibid. 


many to more likely be practical in space than gas systems such as XeCl, which cause 
sharp vibrations when pulsing. However, no "matching" (to the Cs ARF) wavelength 
source of the glass type has so far been identified, and costs of semiconductor diode pumps 
have been high. There are, also, strong interests in diode-pumped lasers for commercial 
appUcations, and for a huge laser for the DoE's Inertial Confinement Fusion program. It is 
the opinion of most experts that a diode-pumped solid state laser wiU be the eventual 
system of preference in space.^^ 

A new MOA indicates the Navy's desire for a continuing Ri&D program on solid 
state lasers for eventual possible use in aircraft or satelUtes. to be conducted jointly with 
DARPA.24 The ongoing DARPA Tactical Airborne Laser Communications (TALC) 
program continues, with Congressional interest, to investigate the use of lasers for tactical, 
possibly two-way cotmnunications between aircraft and submarines, and provides 
oppormnities for test and demonstration of new laser and receiver technologies. 


In retrospect, it would seem that at the time the joint Navy-DARPA program began, 
most of the key technologies, the exdmer laser. Raman converter techniques, the Cs vapor 
atomic resonance filter, the characteristics of optical receivers working against solar 
background.^* and propagation of Ught through clouds and water, were aU known to some 
degree. However, the eUgible lasers appeared to be too large for space use and confidence 
apparentiy had to be built up by tiiose involved in the quantitative characteristics of ARF's. 
An aggressive program plan, outiined by L. Wood in 1979. was greeted widi skepticism 

DARPA initially emphasized the ground based-space minor combination because of 
the DEW motif. On the one hand this may have slowed progress toward a space-based 
system, pushed by the Navy with less funds, and on the otiier may have kept developments 
going which were not possible standing alone. The main technical barriers to a space- 
based laser system were removed when compact discharge excitation of the XeQ laser was 
worked out. and later when the Cs vapor filter characteristics had been improved far 

23 M. White, ibid. 

24 Discussion with Dr. L. Stotts, DARPA, 3/89. 

25 Cf. Donahue, Rcf. 9. p. 2072-2073. 

26 L. Wood, ibid., Ref. 6, and subsequent comments by G. Dinecn. An ad hoc panel of 
Science Board looked into Wood's proposal, ibid, pp. 3740-1. 


enough to reduce the power requirements of the space-based laser system to an acceptable 
level. The GBL approach was removed as a competitor when it was transferred to SDI. 
The program then focussed on reducing risks of the space-based gas laser. 

The DARPA program managers kept high level interest up by a succession of 
successful field demonstrations. SLCSAT and its predecessor were looked on by 
NAVELEX as a "poor horse," in comparison with ELF. and was supported only because 
Congress wanted it But the demonstrations turned out "better than expected" in every test, 
which kept Congress suppUed with ammunition and also maintained some high level Navy 
interest. The persistence of a dedicated NOSC program manager, G. Mooradian, was 
responsible for much of the success of these dcnwnstrations. 

One of the critical Navy arguments for ELF was that it is not "high technology." is 
available now even if only in a quite limited system, cost is not great and it meets a current 
need.2'' Further, SSBN communications requirements have been constantly stated by the 
Navy to be adequately covered by available technology. In any case, the Navy had "closed 
ranks" in the early 1980's in support of ELF. The advantages of the SLCSAT system- 
specificaUy, less restriction on the operating envelope and possibly a sUghtiy faster rate of 
transmission-are not seen by the Navy as outweighing the merits of ELF, which is 
regarded as good enough for now. However, the requirements for communications for 
attack submarines may change in the fiimre. due to such factors as submarine quieting by 
the Soviet Navy. The same threat development also caused the "direct support" SSN 
mission to diminish in attractiveness, and with this, general Navy interest in aircraft- 
submarine communications waned. Because of the change in the threat environment, the 
SLCSAT system defmition, as well as its cost, is correspondingly unclear. 

The weight of expen opinion currently judges the development of an XeQ gas laser 
for a space-based system to be more risky than the development of a new solid state laser 
for space deployment There seems to be confidence that soUd state lasers can perform 
well in space systems. Also, efficient diode-pumped solid state lasers, which are being 
developed by several groups, may provide evenmal cost reductions. A new MOA, initiated 
by the Navy, seems to be prompted by these considerations and provides for a joint effort 
in this direction. TALC can provide an important opportunity to demonstrate this 

27 An "austere" ELF system had IOC Summer 1986. Ref. 5. 


SDI-type developments may eventuaUy improve the general technology of gas 
lasers in space, and increase confidence also in a gas laser for SLCSAT. Also, SDI work 
toward GBL technology for DEW programs may suggest reevaiuation of the ground-based 
laser plus space mirror approach. 

The DARPA expenditures for the space-based laser approach, the demonstrations, 
and the ARF receivers were about $150 milUon at the time of transfer. Expenditures for the 
communications aspects of the specifically GBL approach were difficult to separate out 
clearly from work for the DEW motif. 

The Navy SLCSAT program office estimates that development of a operational 
system could be achieved in the late 1990's, with acceptable risks, but cost estimates vary 
widely from $2 to $30 billion. 





























. 1976 Xa a 



1988 ^ 





A0 1871 (SLC) 

■ DEW 



AO 4011 

AO 5069 





"Y" 1986 







r - 


























:1. \- 



Tank Breaker was undenaken by ARPA in the mid to late i970's in order to 
address deficiencies in man-portable, anti-tank and anti-air weapons. These deficiencies 
were becoming more acute due to advances in armor and other capabilities being fielded by 
the Warsaw Pact forces. Evaluated in a shoot-off in 1987-1988 against several 
competitors, tank breaker technology has been selected for full-scale development by the 
Army as its new noan-portablc anti-tank system, replacing the DRAGON. 


In die caiiy 1970's the Army Infantry Center and the Marine Corps Development 
and Engineering Command identified a number of deficiencies in the DRAGON and 
REDEYE man-portable weapons systems then available to counter tanks and aircraft. A 
problem identified by the Army and Marine Corps study groups was the vulnerability of die 
soldier due to DRAGON'S launch signature. The groups also brought out other 
characteristics that would be desirable, such as being able to "fire and forget" die missUe 
and die capabiUty of launching die missile in confined spaces in urban combat. However, a 
follow-on study by several contractors concluded, in die late 1970's, tiiat die state of die art 
couW not achieve die desired capabilities in a man-portable weapon.^ 

In die early 1970's DARPA set up die ATADS (Anti-Tank, Air Defense System) 
program, to develop a single missile system to counter bodi tanks and die air attack dffeat 
ATADS used a "laser beam rider" (LBR) guidance scheme, widi a flat trajectory. 
However, die Army wanted separate missUe systems for die anti-air and anti-tank missions 
partly because of organization and C^ problems ^ The C^ restraints on launching an air 
defense missile over die battlefield could seriously inhibit anti-tank fire. Apparentiy diere 
were also some NATO discussions about development of two families of weapons, widi 

1 Discussion with Mr. R. Moore, 6/89. 

2 Memo to Dr. Colladay, by J. Entzminger, DARPA. 2/89. 


co-production.3 The Army did undertake a competitive test, for the anti-air role, of the 
ATADS beam rider, against their own infrared (IR) homing system, and selected the ^ 
system. This later became the STINGER. The DARPA anti-air LBR system was later,: 
designated STINGER ALTERNATE. The Army Anti-Armor Command, however,' 
adopted the LBR D ARPA-generated technology for their primary approach to the anti-tank 
.problem.^ More recently, the Anny has used the LBR technology in their line-of-sight 
forward-heavy air defense anti-tank system (ADATS) mounted on the Bradley Fighting 

In the mid 1970's. a number of discussions with DARPA Tactical Technology 
Office (1T0) contractors, and some trials by the Hughes Aircraft Company using 
hcUcoptCTS, led to the conclusion that advances in DARPA-funded focal plane arrays and 
other technologies might offer significant potential for a new man-portable system that 
could achieve the desired miHtaiy characteristics identified by the earlier studies, and also 
deal with threat armor improvcmems. However, due to the relatively recent negative 
smdies by some industrial groups, previously mentioned, DARPA first undertook to define 
and develop an experimental "baseline" system concept that could be tested by^Uie 
Services.^ The concept that resulted embodied (in 1979) a number of DARPA-develop^d 
technologies including: (1) infrared focal plane arrays and associated proce?sin| 
technology, capable of acquisition and tracking of a tank target; (2) a thrust-vector control 
system developed by DARPA to meet low cost objectives, and allowing a "lofted^'mssile 
trajectory to attack the top, thinner tank armor, (3) an advanced shaped-charge warhead. A 
Smokeless, off-the-shelf propellant aUowed a low-velocity missile launch with low 
signature and permitting operation in confined spaces. This new systems concept, using 
the infrared focal plane arrays, departed significantiy from DARPA's earlier LBR 
approach, which the Army Anti- Armor Conunand had already adopted. The concept 
envisaged a "lock-on before launch" mode of operation, with the soldier being able to sight 
tiie target through the missUe acquisition optics. Once locked on and fired the missile was 
on its own in a Tire and forget" mode. LSI processors and advanced algorithms permitted 
different modes of guidance in earUer and later stages of the missUe flight. The overall 

3 Discussion with Mr, R. Moore. 6/89. The problems of establishing a NATO program apparenUy were 

not resolved. 
^ Dr. J. Entzminger, ibid. 
^ OTE Report to Congress, FY 1988, p. 111-13. 
6 A.0. 3239 of 3/76, "Fire and Forget Science and Technology." 


system was lightweight, about 35 lb, to meet portability objectives. There was also 
potential for system growth to allow distant launch fipom helicopters. 

This concept, illustrated in Fig. 1, became "Tank Breaker," a coordinated program 
with the Army's Intermediate Man-Portable Anti-Armor Weapons Systems (IMAAWS) 
program, and the Marine Corps. The first Tank Breaker program was to have two phases, 
the first phase (12-months) starting in 1980 to demonstrate component technologies and 
-their integration, and the second phase (24.months) for missile system and warhead 

There were four industrial groups involved, following two different approaches. 
The progress was rapid in the first phase, demonstrating all the critical technologies and die 
superiority of the Texas Instruments-Hughes approach. As a result the Army cancelled its 
IMAAWS program plans. In fact, significant advances in the state of the art of focal plane 
array seekers and trackers had been achieved and demonstrated to woric in this first phase, 
and further questions remained only in the selection of seeker wavelengths and die design 
of the tracking and guidance system. 

By the end of the second phase, more of the key questions were resolved and 
several successful flight-test demonstrations had been conducted. In accordance with the 
DARPA-Army agreement MOA, Army took continuing responsibility, in 1979, under its 
new Anti-Armor Weapons Systems-Man Portable (AAWS-M) program. For nearly four 
years, however, further Army action was held in abeyance, apparently due to controversy 
regarding the technical risks, costs, and operational utility relative to approaches based on 
LBR designs, which were still favored by some Army developmental groups. Because of 
continuing pressure by the Army and Marine Corps user communities, however, the Army 
decided in the late 1980's to have a "shoot-off* between the contractors. A new LBR 
design was involved in this test as were two vendors of Tank Breaker with differing 
designs. Evaluation of the results led to selection of the Texas Instruments Tank Breaker 
design based on the DARPA-developed technology. DSARC Milestone H review was 
scheduled for early 1989,8 and approval was given, in June 1989, for full-scale 
development pending additional operational tests to compare with an upgraded DRAGON. 

A.0. 3974, "Anti-Aimor Assault Missile" of 3/80. 
8 Discussion with M. Bair, IDA, 7/89. 


Some continuing concerns also have been expressed about the costs and reliabUity of 
sophisticated "fire and forget" technology.^ 


Tank Breaker represents a timely interaction of technologies to meet a pressing and 
fairly specific statement of needs by miUtary user communities. Tank Breaker's approach 
to meeting these needs did not imply a radically different mode of operations, but would 
aUow a large improvement in infantry anti,-tank capabUities by allowing much more 
flexibiUty and providing reduced vulnerability. The early industry reaction to the need 
statement was that meeting it would be beyond the state of the art However, the potential 
of the new DARPA-developed focal plane array technologies as a key element of a system 
to meet diese needs was indicated by industry initiatives. DARPA undertook further 
development and integration of this and several other technologies involved in such a 
system. Because of the complexity of the technology this was seen by some as a fairly 
risky endeavor. Throughout, there was strong support from the user conmiunity, and 
resistance from some of the Service development groups. 

Part of this resistance apparently stenmied from what could be regarded as a 
previous successful transfer of DARPA LBR technology, which Army's MICOM 
embodied in their prcfeired approach to an anti-tank weapon. The LBR technology which 
had been developed by DARPA under the earUer ATADS program was aimed at a soldier 
portable weapon for both anti-air and anti-tank use. The Army did not accept this common 
missUe appioach which could not be optimized technically for both missions. Although the 
anti-air LBR lost in competition to the IR-guided STINGER. MICOM did continue woric 
on the LBR for the anti-tank mission and ATADS provided some of the missile technology 
that was integrated by DARPA into Tank Breaks. The LBR has now been adopted by the 
Anny for their forward air defense system mounted on the Bradley Vehicle. 

Part of the Army's resistance also came from concerns regarding the costs and 
reliabiUty of tiie sophisticated Tank Breaker technology. However, since Tank Breaker 
(now AAWS-M) was closer to the users' desiderata, it had their support. The shoot-off 
test eventually conducted by the Army seems to have senled the problem of selecting 
between advanced options. However, a recent modification of the existing DRAGON 
provides a low-cost option which is to be tested against the AAWS-M. 

9 Discussion with M. Taylor, IDA. 7/89. 


From project records, D ARPA outlay for Tank Breaker itself appears to have been 
about $35 mUUon, which does not include earlier development of focal plane arrays or 
other technologies eventually incorporated. Expected AAWS-M procurement expenditures 
are about $2.8 billion. i° 

1 0 DMS Market Intelligence Report, Missiles, AAWS-M, Jane's 1988. 





















L - - 


I I 
I I 



■ ■ AAWSM 














7-31-89-1 4M 




In 1973 a joint Army-DARPA program constructed a high- velocity, rapid-fire 
75 mm gun of novel design, incorporating several emerging advances in ammunition, 
propeUant and fire control technologies. This program was soon expanded to encompass 
construction of two lightweight test-bed vehicle gun combinations, HIMAG (High 
Mancuveiability-Gun) and HS VT/L High SurvivabiUty Vehicle Technology (Light). After 
successful gun trials, the Army took full responsibility in 1977 for an accelerated 
HIMAG/HSVTA- test-bed program. Thorough test, evaluations and analysis indicated 
feasibility and generated for the first time a quantitative data base and modeling 
methodology relating performance to weight and cost of gun-vehicle combinations. 
Satisfactory performance against threats in the mid 1980's apparcndy demanded weights 
higher than the Army's air tianspon limits. 


In 1973 DARPA began a joint program with Ac Army aimed at a lightweight, high 
velocity (HV) cannon for use against medium to heavy tanks and low performance 
aiiciaft.1 Parts of the motif for this program came from carUer DARPA studies of an "anti- 
tank machine gun" to deal with the large numbers of targets expected on NATO battiefields, 
the developing concepts within the Army of a completely air-transportable division, and 
also from the Marine Corps icquiicments for a heUcoptcr-transportablc "mobile protected 
weapon system," or Ught tank. Partly also it was felt that a Ught, agile vehicle carrying a 
HV cannon might have high survivabiUty and effectiveness on future battiefields with a 
corresponding impact on tactics. A 75mm caUbcr was chosen for demonstration of a 
hypcrvelociiy smootii-borc, lightweight cannon, to be capable of rapid, highly accurate, 
automatic burst firc.2 Initially. Uquid propellants were investigated but soUds were soon 

Testimony of Maj. TeircU G. Covington, p. 3067, in DoD Authorization Hearings for FY 1980, 
Committee on Aimed Services, U.S. Senate, 96ih Congress, 1st Session, Part 6. 
A.O. #2447, 2/73, "75mm Liquid Propellant Gun." 


chosen as more mature technology. The medium caliber anti-aimor automatic canno^^ 
(MCAAC) was to be designed for low recoil, and also to have new "kinedc^energyiK' 
penetrating ammuninon. 

In 1973, also, joint studies began by DARPA and the Army, in a new advanced 
combat-vehicle technology (ACVT) program, to investigate perfonnance paramciq^tdiat 
could be achieved by integrating sevcial emerging technologies, including the 75mm gu^. 
advanced fire control and new Ughtweight armor, into vehicles with a fiiU-up weight in tfic 
range of 12 to 40 tons. In 1975 DARPA and the Army jointly funded construction pf .• 
HIMAG in the upper (40 ton) weight range 3 The HIMAG was envisaged not a^ 
prototype, but as a test bed which would be modified almost continuously to obtam 
performance data at dif fexent weights and costs. 

Specifically, the HIMAG System: 

basically was fabricated to provide variability and to specifically address 
mobility, agility, and association with horsepower per ton and suspension 
systems, and also to address fire control system options. 

Specifically, that variabiUty includes being able to vary the power, the weight of the 
system, the running gear combinations, the suspension system levels, the firing 
system of automatic, semiautomatic and or single shot firing with the automatic 
cannon, and a fire control system which can be varied in, sophistication frpm a ^ 
simple fire control iron sight up tiiough a closed loop, distance sensing, thCTmal 
imaging, automatic cracking fire control system.^ 

The 75mm cannon was designed by Stoner (who had designed the AR-15 
predecessor to the M16 rifle) and produced by ARES and had a very successful feasibility 
demonstration in 1975, firing from a fixed platform. This led to an acceleratiop of the 
75rmn program, and die fabrication of advanced ammunition, which included a compact 
"telescoped" APFSD (armor-piercing, fin-stabilized, discording SABOT) round with a 
long rod kinetic energy penetrator. In die fall of 1976 the Marine Corps jbined die 
DARPA-Aimy program. Further successful trials were held in 1977, demonstrating 
penetration of diick armor at long range, acceptable shot dispersion and gun corrosion, and 
high rates of fire. The results aroused considerable enthusiasm in Congress, which 
appropriated $1 IM extra, and in the Army Chief of Staff, Gen. Rogers, who moved up the 
IOC for the system to 1985 from 1990. In 1977 the Advanced Combat Vehicles 

3 A.O. 3130, HIMAG, 10/75. 
^ Covington, ibid. 


Technology (ACVT) Program Office was formed directly under the Chief of Staff of the 
Army, who accepted full responsibility for further development and for expansion of the 
program to meet Marine Corps objectives. DARPA continued support for selected high- 
risk technology aspects, particularly in fiie control, since the 1977 tests showed some 
weaknesses in this area. 

As one of the ACVT's first activities, the Army's Tank R&D command began 
construction of the HSVT/L test-bed. in the 15.029-ton range, and carrying the 75mm 
MC/AAAC gun (See Figure 1). 

As described by the program manager, who moved to ACVT from DARPA, 

The HSTV/L brings together in the 15- to 20-ton class test-bed a number of 
technology options for examination. These include the hunter-killer fire control 
which is represented by two independent sight heads. In this case one member of 
the crew may select, identify, and acquire while the other sight system is dedicated 
in conjunction with tfic gun to firing or engaging against a previously selected 

And regarding objective, 

The objective is higher targeting and servicing rate, in the functions of an automatic 
cannon, in combination with a fire control system which allows us to overlay the 
two actions of identifying, acquiring, and selecting targets with the acmal 
engagement process.^ 

Tests of the HIMAG and HSVT/L began in 1978, with the 75mm gun firing on the 
run while moving over different types of terrain, and using several different types of fire 
control systems. Tests of "full up test systems" (PUTS) continued through 1980. 
Figure 1 shows one such system. Recognizing that the number of acmal tests would be 
limited, provision was made for simulations and modeling. The statistical data and 
simulation methodologies, developed partiy with DARPA support, were judged sufficient 
to support an evaluation of HIMAG and HSVT/L that year by AARADCOM. This 
evaluation judged firing performance to have been moderately successful, while kientifying 
a number of desirable improvements, notably in infrared systems for fire control, and also 
recommended work with a higher caUber cannon, 90 mm or more, to deal widi future 
threats. Studies of a 90-mm cannon-vehicle using Uie methodologies developed were 

5 Covington, ibid. 
^ Covington, ibid. 


A number of foUow-on studies by Army doctrine and in infantry commands were 
conducted in the early 1980's. to define systems and describe trade-offs. The conclusions 
pointed to the feasibiUty of a 75mm gun-vehicle combination in the 21-ton range. DSARC 
was anticipated in 1987.*' 

As this date approached, however, it appeared increasingly difficult to meet the 
requirements for air transport weights with acceptable performance characteristics. The 
growing appreciation in the early 1980's of improvements in Pact armor also impUed a 
need for a higher caHber gun and heavier anmiunition, also discouraging further steps 
towards acquisition. The Army's present ADATS (Air Defense And Tank Systems) 
approach involves laser-beam-riding missiles mounted on the Bradley Fighting Vehicles 

The Marine Corps, however, with different threat priorities, continued interest 
through 1986 in the potential of the lightweight 75mm gun for use on its LAV high 
armored vehicle.^ 


HIMAG appears to have originated in a joint DARPA- Army program towards a 
75mm, rapid-fire gun for use on Ughtweight combat vehicles. The 75mm gun system was 
a new design and was to incorporate a number of emerging propellani and ammunition 
technologies. However, one of these technologies which was pushed initially, the Uquid 
propeUants. was evenmally abandoned since the technology proved insufficiently manire. 

Early successful trials with the 75mm gun led to program expansion to construct 
HIMAG. a test-bed vehicle to carry the gun and have the latest armor, engine and fire 
control technologies. Further success with static firing of the 75mm gun led to enthusiastic 
acceptance of the program by top levels in tiic Army in 1977 and extra support tiiat year 
from Congress. 

7 "Medium Caliber Ami-Armor Automatic Cannon Programs," (XJ), Final Report, VoK 1. 
USARRAOCOMM 1982, P. 5 (Confidential) Unclassified excerpts have been made from this report 

8 DoD OT&E Report to Congress for FY 1988, p. 1 1 1- 13. 

9 Jane's Armor and Artillery, 1987. p. 870. 


Figure 1. HIMAG/HSVT-L Tank 

TTie HIMAG and the later and Ughter HSVT/L were intended to be test-beds which 
would be modified and evaluated in the course of field trials to judge the range of 
capabiHties provided by emerging technologies. HIMAG/HSVT/L fulfilled the test-bed 


role, providing for the first time a data base and methodology from which adequate 
decisions could be made regarding technical performance military utility and 
transportability. Generally, the technical performance seemed satisfactory, except for IR 
fire control. By 1980, however, there were some early indications of Pact armor 
improvements, leading to recommendations for a larger gun. The test results and 
associated studies indicated, as time went on, diat HIMAG/HSVT/L would not be able to 
meet the maximum weight limits set by air transport, with acceptable performance, 
especially when taking into account the threat expected for Army priority missions. The 
Marine Corps, with different priorities, continued interest in the Ughtweight gun's potential 
for several more years. 

The DARPA Ughtweight gun and HIMAG program appears to have been a success 
in that relatively quick transfer took place to the Army, with full backing by Congress. The 
decisive factor for the Army's decision not to proceed after about 1982 seems to have been 
the minimum weight required to deal with advances in the threat, which were apparenUy 
not fuUy anticipated until after the transfer had taken place. The HIMAG experience and 
data, however, appear to have given the Army for the first time a quantitative basis and 
metiiod of evaluation of trade-offs of vehicle, gun. and fire control characteristics against a 
given threat 

DARPA outlays, from project records, were about $25 million to the time of 
transfer. About $22 miUion more was spent by DARPA on HIMAG after the transfer. 












90 MM? 










The potential of mini remotely piloted vehicles (RPVs), integrating new sensors 
and C3 technologies with that of improved model airplanes, was demonstrated by ARPA's 
PRAEIRE and CALERE in the early 1970's. These mini-RPVs affected the IsraeU 
developments of RPVs which were used in the 1982 engagement with Syria, and 
influenced the Army in its AQUILA program. The U.S. Navy and Marine Corps have 
acquired IsraeU MASTIFF and PIONEER RPVs for operational tests and use. 


Attempts to use unmanned, remotely controlled air vehicles go back to about the 
time of WW V In the late 1920's remotely controlled aircraft were built in the U.K, and 
U.S.. and used mainly as target drones and guided bombs. Between the wars there were 
some industrial efforts to construct drones for target practice, and these were greatiy 
expanded in WW H. In WW H. all the U.S. miUtary services also made attempts to use 
radio-controUed aircraft for special missions, some involving television cameras in the 
vehicles. Similar efforts continued through the Korean War. 

In the mid 1950's, the U.S. Army undertook a program to develop several types of 
what were then called radio-controlled drones, to be used for a variety of purposes, 
including reconnaissance, target acquisition, strike, and electronic warfare.^ Typical 
weight for these drones was about 450 lb, and the fUght duration approximately one-half 
hour. The vehicles for some of these missions were envisaged to have quite low costs. 
However, by tiie early 1960's, and after expenditures of about $800 million, all but one of 
the projects had been cancelled because of complexity and high costs. Besides the 

1 Some early history of RPVs is recounted in War Without Men. Pergamon-Brassey, 1986. p. 31 ff. 

2 John Kreis. "Background of United States UAV Activity," IDA. unpubUshed ms. and DSB Smnmer 
Study, on Remotely Piloted Vehicles. 1971. Appendix A (Classified). Unclassified excerpts have been 
made, in this article, from this and other classified reports cited. 


tendency to increased complexity, some of the Foblems that appeared in this early work 
reappeared in later efforts, notably propulsion engine and communication- navigatiorf . 
systems reUabiUty, In 1964, dte Army abandoned most of their program and the Chief of ^ 
Staff stated that the Army would depend on the Air Force for many of the missions anM 
information which they had hoped to obtain from the radio-controlled drone. In 1965, and^ 
apparentiy in response to pressures of the Vietnam War. the Army declared their surviving t 
drone (the SD-1), which had been used for training, "operational" despite its kno^ 
deficiencies. The SD-l. redesignated the USD-5. was not used for long, however, and by^: 
1966 the Army was no longer active in tiie remotely piloted vehicle area, except for 
conceptual studies.^ 

After the Cuban missile' crisis in the early 1960's, the U.S. Air Force began the 
BIG SAFARI program, a large program including an effort to develop a substitute for the 
' U-2 for reconnaissance in heavUy defended areas. This led to a modification of the Ryan 
Firebee. previously used as a target drone, to produce die first jet propelled drone 
reconnaissance vehicle, which had operational flights over China in 1963i^ TheiFirebee 
vehicles, designated AQM's and BQM's, were further developed to reach progressively 
higher altimdes to improve survivabiUty. These Air Force drones, while much smaller than 
a manned aircraft, could still accommodate sizeable payloads. These were launcjied from a 
"mother" aircraft in the successful BUFFALO HUNTER reconnaissance effort m Vietnam. 
Some of the Air Force drones were modified in 1964 for use at low altitudes in Viemam. 
This experience and threat inteUigence led to a reappraisal of survivabiUty and to eventual 
drone redesign favoring very-low-altimde, high-speed runs. Several hundred of these low- 
altitude drones were obtained and used mainly for reconnaissance and electronic warfare 
missions in Viemam, with over 3500 flights and considerable success.^ Considerable 
operational experimentation went on to solve die navigation problems, eventually largely 
overcome by use of TV systems on the drones. In the mid 1970's the Air Force further^ 
modified several of their drones to gain a capabiUty to destroy air defense radars aa^ other, 
targets, using TV-guided missiles such as Maverick.^ In retrospect, the Air Force felt tha^^ 
while successful, their Viemam drones had high support costs, which discouraged follow- 

3 Address by Brig. Gen. W.H. Vinison, "Army Perspecuve on the Use of Surveillance and Targeang 
RPVs," in Proceedings of the Symposium on Remotely Piloted Vehicles, Nauonal Bureau ot 
Standards^ May- June 1972, p. 293 (Classified). 

^ War Without Men, ibid., p. 31. 

5 Ibid. 

6 John Kreis, ibid. 


on efforts. These were high costs in peacetime when the alternative costly manned aircraft 
were not being attrited^ 

In 1959 the U.S. Navy began development of the drone-anti-submarine helicopter 
(DASH) the first helicopter RPV system, mainly to enhance the capability of small vessels 
for Anti-Submaiine Warfare (ASW), However, due to interfering electromagnetic signals 
aboaid these ships. DASH proved difficult to control. The Navy cvenmally abandoned the 
DASH program in 1970, but not before several of the heUcopters were equipped with low- 
light-lcvcl TV systems, renamed SNOOPY, and used at night to assist the Marines in 

In the late 1960's, and apparently in response to a "Zap channel" request from 
ODDR&E, ARPA's Advanced Sensor Office (ASO) undertook to improve SNOOPY.* 
ARPA added a number of new systems to the DASH, which had considerable payload 
capability, making two experimental systems called NITE PANTHER and NTTE 
GAZELLE.' The payloads at various times included, besides communications and 
guidance packages and day- and low-light-lcvel TV, a moving target indicator (MTI) radar, 
a hypcrvelocity gun, a laser designator-rocket system and a variety of other weapons. The 
TV's were of both low and high resolution variety, with stabilized optics for the high 
resolution system. The NTTE PANTHER was ^aientiy used first in Vietnam, mainly for 
tests and demonstration of remote target acquisition capaWlity witii accuracy sufficient for 
fire control. NITE GAZELLE was intended to be a standoff, precision strike system. 
Both of these were used successfully for training and operational nodssions in Vietnam until 
the early 1970's, but were plagued for some time by mechanical reliability 

The success of these helicopter systems and the need for greater range for the 
RPVs led the ASO to the concept of the "extended battlefield," using the tethered balloon- 
borne systems: EGYPTIAN GOOSE, with an MTI radar for tracking, and the 
GRAND VIEW for TV-bandwidth communications.i^ a number of tests of the NTTE 
GAZELLE extended range system were conducted in the early 1970's at NeUis Air Force 

7 Hearings on National Defense Authorization for FY 1988-1989, HR 1748. Title I, p. 208, and 
conununicsuion from Dr. A. Flax, IDA 2/90. 

8 "SNOOPY-Zap Channel," AO 1162. 2/68. The Zap Channel was a quick reaction mechanism by 
which ARPA would respond to urgent DDR&E requests for Vietnam. 

^ AO 1200 of 3/58. NTEE PANTHER and NTTE GAZELLE. 

10 Discussion with J. Goodwyn, 3/89. TTie mechanical problems were eventually solved. 

11 EGYPTIAN GCOSE was the predecessor fiat POCKET VETO, described in Chapter XVn. 


Base, demonstrating the capability to find and designate targets for attack over 100 nmi 
ranges.i2 The payload in KITE GAZELLE, used in these trials, included a rocket with a 
laser angular rate seeker which was the beginning of work by Martin Marietta which led 
eventually to the seeker used in the Army's COPPERHEAD laser-guided inunition.i3 The 
NiTE GAZELLE was apparenUy regarded as an expensive system, since the first one cost 
over $10 miUion to develop, and its reputation for reUabiUty difficulties discouraged large 
scale use.^^ 

ARPA intensified efforts, in the early 1970's, toward development of Ughter, more 
compact, higher performance and lower cost electrooptical systems for use in Viemam, 
both on die ground and in the RPV's. 

Also, in the early 1970*s, new technological advances in composite materials, 
sensors, navigation, and vehicle design and propulsion, together with an increased 
appreciation of the air defense threat, led to new DoD interest in die possibilities for use of 
RPVs. In 1970, DDR&E established a special R&D initiative in tfiis area.^^ A number of 
smdics and symposia were held in die 197 1-1972 period to help determine die state of the 
art and define directions for an intensified DoD program.i6 In particular, a 1971 Defense 
Science Board (DSB) panel on RPV's outlined a set of desirable characteristics based panly 
on extensions of model airplane technology, and on the previous experience widi AF 
drones and ARPA's NTTE GAZEU-E.^^ The DSB's Ust of payload characteristics was 
similar to tiiosc for NTTE GAZELLE, but die subsystems involved had to be much Ughter 
and smaller to fit into die mini-RPV concept suggested Much of die needed technology, 
die DSB noted, was available, but fiirdier research was needed on lightweight infrared aR) 
sensors and on C? problems. In contrast to die Vietnam experience widi drones, die DSB 
felt diat RPV costs could be kept low. The mini-RPV concept oudined by die DSB was 
given die acronym RPOADS (Remotely Piloted Observation and Designation System), 
which was used by die Army for dieir follow-on RPV program. At an early stage of its 

1 2 "Advanced Standoff Weapon and Sensor System." Vol. 1. RCA Service Company, 15 June 1972. 

13 Discussion with R. Whalen, Martin Marietta. 12/89. 
J. Goodwyn, ibid. 

^ 5 NBS Symposium, ibid., keynote address by liD. Benington, p.3. 

16 "Remotely Piloted Vehicles, An Idea Whose Time Has Come." Report of the ^"S/. Pj. 
AFSC/Rand Symposium of May-July 1970; "Report of the Panel on Remotely I^J?«fi Jc*»»^J«' 
DSB Summer Study, 1971; NBS Symposium 1972. Also. Battelle conducted a special study of the 
RPV/State of the Art for ARPA in eariy 1971. AU these reports are classified. 

1 Defense Science Board study, ibid. 


RPOADS program the Anny requested ARPA to conduct a number of trials of the NTTE 
GAZELLE system at NelUs AFB, which demonstrated successful designation of fixed and 
moving targets.^^ jn 1972 also, the Army Chief of Staff expressed dissatisfaction with the 
respcHisc of the Air Force to the Anny request for battlefield assistance after the Army RPV 
program was cancelled in the mid 1960's. 

In the early 1970's, Israel conducted intensive smdies of the possible use of RPV's 
in engagements against the heavy air defenses being set up by the Egyptians and other 
possible enemies. (The possibilities of RPVs in this theater were also discussed briefly in 
the DSB 1971 report) Apparently Israel was able, about this time, to obtain some of the 
USAF-type reconnaissance and target drones from the U.S., which they subsequently 
modified. 19 In their 1973 war these Israeli RPVs were used quite successfully. 

In the early 1970's also, apparently during one of the briefings given by ARPA to 
Dr. John Foster, then DDR&E Director and also a model airplane enthusiast, he 
recommended that the ARPA program should not continue witii expensive and compUcated 
heUcopters such as NTTE GAZELLE but should be oriented toward use of Ughtweight, 
rugged, inexpensive noodcl airplane technology .2° 

The ARPA mini-RPV program began shortly thereafter, in early 1972, as an effort 
toward the type of lightweight, compact, low-cost sensor/laser target designation system 
that had been recommended by Dr. Foster and the DSB.21 The resulting PHELCO-FORD 
RPV had exchangeable modular payloads, the RPV carrying the daytime TV-laser target 
designator configuration called PRAEIRE, and the same RPV carrying a Ughtweight FLIR 
and laser target designator combination, called CALERE. The propulsion system was an 
adaptation of an engine tiiat had been used in lawn mowers. The radio command was also 
adapted fiom one commercially avaUable. and was operated by a pilot and a sensor 
controller. Vehicle stabilization was provided initially by an electrical field sensing system 
developed by John Hopkins AppUcd Physics Uboratory; later, gyro stabilization was 
apparentiy used.22 Optical stabilization was provided for the high resolution TV, and die 
laser designation systems used the same optical sighting train as die TV, as had been done 

1 8 Remotely Piloted Vehicle Laser Target Destination Tests, US. Army ECOM Technical Report 4054. 
November 1972. 

19 J.Kreis,ibi(l 

20 Discussion with Mr. James Goodwyn, DARPA, 3/88. 

21 AO 2047 "Zoom" FLIR." 1/72 and AO 2056. "Mini Laser-Sensor Designation System." 1^2. 

22 "World Unmanned Aircraft," by K. Munson. Jane's, 1988, p. 155. 


in NTTE GAZELLE. PRAEIRE I, the first of two versions produced under the ARPA 
program, weighed 75 lb and had a 28 lb payload and a two-hour flight time.23 It was 
described as an austere, low-cost system, with a cost estimate, in mass production, of 
$10,000/copy.24 The first flight of PRAEIRE I occurred in 1973 after a joint ARPA-Anny 
program had been started.25 However, there were some difficulties with pcrfoimance of 
the CALERE IR payload, requiring further development.^^ 

The Army's effort in response to the DoD initiative included, besides the joint 
program with ARPA, trials of several other types of available mini RPVs in a program 
intended to gain a better detennination of requirements, caUed "little r."27 Part of the "littie 
r" program also was a phased developmental effort of an entire RP V system, together with 
ground control and support, which led to the Lockheed AQUILA. beginning in late 1974. 

During the 1972-1975 period, ARPA produced PRAEIRE H and CALERE H, 
again buUi by Ford, based parUy on the experience with the previous vehicles, and pardy to 
reduce radar and IR signatures. Sensors and propulsion were also improved, with fUght 
ume capabiUty extended to nearly six hours. The extended range vehicle PRAEIRE H B 
had nearly twice the weight of PRAEIRE 1.28 An electronic warfare payload was also 
developed. CALERE m was also produced, including a new, lighter FLIR-laser target 
designator ccnnbination. 

In late 1974, a joint ARPA-Army effon commenced to develop an integrated 
communication-navigation system.29 a UtUe later a PRAEIRE RPV successfully 
demonstrated the capability of designating a tank target for the Army's COPPERHEAD 
cannon-launched guided projectile.^^ 

TTic Navy, besides its DASH program and its use for SNOOPY activity in Vietnam 
also conducted trials of Air Force drones in 1969 and 1970 which indicated feasibility of 

23 Munson, ibid. 

24 Hearings before the Committee on Armed Services. HOR. 1976 and 76T Appiopnauons. 94ih 
Congress. Isi Session. Testimony of K. Kresa, p. 3973. 

25 Hearings, ibid. Testimony of Brig. Gen. Dickinson, p. 3985. 

26 Hearings, ibid. Testimony of K. Kresa, p. 3973. 

27 Brig. Gen. Dickinson, ibid. 

28 Jane's, ibid. 

29 -Integrated Communication Navigation System," AO 2922 of 1 1/74. 

30 -PRAEIRE Mini RPV Laser Target Designation System." Signal. Feb. 1976, p. 70. 


operating from carriers.3i In 1973, with a better picture of its requirements, the Navy 
joined DARPA in a program to develop an RPV capable of being operated from small 
ships.32 This joint effon produced and tested the Teledyne STAR, in a one-year effort. 
Considerable difficulty was experienced, as anticipated, with shipboard recovery.33 

Until the early 1970's the Air Force had not been involved with mini-RPV's.^^ In 
1973, DARPA began development of the AEQUARE mini-RPV, capable of being launched 
from an aircraft, for target designation in a heavily defended area. After several 
demonstrations, die Air Force had a brief follow-on program which ended in 1976.35 

In the early 1970's also, DARPA and the Air Force conducted a joint program to 
develop an expendable mini-RPV, capable of loitering and attack, called AXILLARY.36 
The Air Force followed up AXILLARY to a limited extent but has apparently favored the 
TACrr RAINBOW loiter-capable, air-launched guided missile, classified until recently, for 
the same mission.^^ 

By 1977 DARPA's early mini-RPV effort had nearly concluded In 1977 also, 
Israel obtained DoD apprc>val to buy several PRAEIRE H B systems 38 The laser target 
designation payload may not have been included in the package sold. Israel went on to 
develop its MASTIFF RPV. later the SCOUT and more recenUy the PIONEER, While not 
identical to PRAEIRE II and incorporating independent Israeli research, these Israeli 
developments appear to have been influenced by the DARPA developed technology. A 
photo of PRAEIRE IIB is shown in Fig. 1. 

During the mid 1970's, the Army's AQUH-A program continued, reaching full- 
scale development in 1979. After a number of difficulties with engine reliability, recovery 
procedures, and C3 technology had been overcome, AQUILA had a series of successful 
tests in the mid 1980's.39 AQUILA's weight, however, had grown to 250 lb together witii 

3 1 Hearings* ibid., testimony of CapL Hill, p. 3292. 

32 "Ship Deployablc Tactical RPVs." AO 2674, of 11/73. 

33 CapL Hill, ibid. 

34 Hearings, ibid., testimony of Brig. Gen. Hodnette, p. 3997. 
3^ Munson. ibid., p. 165. 

36 -DefenseSiqipression." AO 2456 of 11/73 

37 Cf., e.g., J.D. Morocco, "Development Test of Tacit Rainbow on Navy A6 Set to Begin Next Week," 
in Aviation Week, July 3, 1989, p. 21. 

38 Munson. ibid., p. 55. 

39 DoD OT&E Report to Congress. FY 1988. p. in-2. 


a $1 million cost as a result of greater capabiUty and more stringent requirements. For 
example, the RPV's operations concept, originally to assist artillery battalions, had been 
extended by 1984 to use by an entire division for a variety of purposes, with corresponding 
additions to the payload.^o Target tracking during jinking maneuvers to survive the 
battlefield were deemed necessary, and anti-jamming requirements for use in the NATO 
theater were difficult to meet and had increased the size and weight of the key Modular 
Integrated Communications Operations and Navigation System (MICNS). Test of 
AQUILA began in November 1986 with the TV payload only, because of continuing 
difficulties with the IR sensor .^i The AQUILA program was cancelled in FY 1988 after 
Congress had refused to fund procurement and cstabUshcd the joint RPV, now UAV 
Program Office (UAV SPO) in DoD. However, the Army apparentiy is planning a new 
RPV program in conjunction with the UAV SPO.*^ 

Figure 1. PRAEIRE ilB Minl-RPV 

Source: WorldUnmanned Aircraft, p. 155 

40 -Results of Forthcoming Critical Tests are Needed to Confmn Aimy RPVs Readiness for Production." 
GAO Report- GAO/NSIAD 84-72, April 1984, p. 13. 

41 OTE, ibid. 

42 J. Kreis, ibid. 


The successful Israeli use of mini RPV's against Syrian air defenses in 1982, their 
tracking of Gen. Kelley of the Marines in Beirut by a RPV when he moved about the area, 
and the Navy's experience in Lebanon in the early 1980's, particularly the loss of an 
aircraft, led Secretary of the Navy John Lehman to order in 1985 that the Navy obtain a 
RPV reconnaissance and gunfire direction capability as soon as possible, using available, 
proven RPV systems/^ In response, the Navy and Marine Corps rapidly acquired first the 
IsraeU MASTIFF, and more recently the PIONEER. The Navy has apparently successfully 
operated and modified the PIONEERs for use from several types of ships and had 
evaluated the PIONEER in operational exercises.^ 

In the 1970's, the Air Force had the COMPASS COPE program for a long- 
endurance high-altitude RPV to replace the U-2. After a shon-time, the Air Force reduced 
funding for COMPASS COPE, citing high cost and lack of clear mission objectives. In 
1983. DARPA undertook a long endurance RPV program, AMBER, taking advantage of 
new advances in materials, computers, propulsion, and sensor capabilities/^ While still 
emphasizing endurance and survivability, the AMBER program became a joint effort with 
the Army and Navy and has produced a variety of RPVs of different sizes for use at high 
and medium altitudes, some of which arc capable of autonomous, "intelligent" activity. 
DARPA encouraged innovative industry participation in the AMBER program. DARPA 
transfeired AMBER technology to Uie Navy and the UAV SPO in 1988. Figure 2 shows 
one of die AMBER vehicles. Botii die AMBER high-altitude RPV and die CONDOR, 
produced by Boeing Company and supported recentiy by DARPA. have set new records of 
altitude and endurance for piopeller-driven aircraft The CONDOR, shown in Figure 3. is 
a large RPV witii a wing span of 200 ft Operational tests with CONDOR have been 
pcrfoimcd witii die Navy to help develop mission concepts and test sensor suites. 


ARPA's NTTE GAZEIXE helicopter RPV program, and a suggestion by DDR&E 
and DSB to adapt its technology for integration with model airplane dimensions, apparentiy 
led to ARPA's mini-RPV programs. Construction and demonstration of die 

*3 j.Kreis.ibid. 

OTE Report to Congress FY 1987. p. IV-71. 
45 A04981of 12/83 AMBER. 


Figure 2. AMBER 500 Flight hrs; 38 hrs. Endurance; 27,800 Ft Photo 

Souxce: From Leading Edge, Inc. 


Figure 3. Condor 
Source: Boeing Company Advanced Systems 


PRAEIRE and CALERE RPVs showed the Services, and the IsracUs what could be 
STSpA-s success tnay have been nuunly in this titnely mcet«g of the tn.n.RPV 

ARPA technology demonstrations. Other ARPA n^ni-RPV prognuns with the Air Fo,.e 
rt^et.^seltohaveledtoService|«ogn«»s.i*shortUves Ho^v«,*eI^eU 
MASTIFF SCOUT and PIONEER seem to be mote direct denvaaves of the ARPA 


In the nud 1970's connnents were made by Navy and Army program manag«.. 
that miHtarized mini-RPVs are not simple modifications of model airplane technology, but 

Tsfu, the technology of a weapons system.^ T-<^-«^f ~ ^^^^ ^ 
expendable vehicles, more nearly the original mini-RPV motrf. and more complex, 
survivable RPVs or high cost manned aircraft are still being debated. 

The AQUILA development led to a complex, heavy, and costly RPV. which was 
recently canceUed. The Army's reasons for .he AQUILA history are bas.d par^y on 
rrg«trequirementsfbrandj«nc3pabiU,y.oopera.ei«theNAT0th^^ ParUy^ 

Zt2 Change in operadonal concept, in midstream, from what was «^^a «^ 
designator for a battaUon's smart weapons, .othis plus a more complex mteU.gence 
^^S^alelectronicwarfaredevice^division-wi^ '^^1^^^ 
Tevoludon occurred, app^dy. in .he Army's earlier prtgnm. .„ ^^^^^^ 
RPV functions seem to have been separated again in more recent Army concepts. 
Sunder the aegis of ti« DoD joint RPV (now UAV) prograrn office, ^up by 
Igressio^d directive in the late 198as. and is appamntiy ^^^^^^^^^^ 
to take the place of AQUILA. Use of an RPV in conjunction with COPPERHEAD was fo 
a time an important driving force for continued Amiy RPV efforts. 

Piloted Vehicles," 1977. p. 24. 

47 GAO Report, ibid., p. 6 .,go7 HR 4428 Tide I. Testimony of Gen. Knudson. 

48 Hearings. Defense Audiorizauon Act of 1987. H.R. 44^5, uuc 
p. 287. 


The Israeli RPV success in their 1982 engagements, which has had major impact 
worldwide, can be credited, partly, to the development of the DARPA technology they 
acquired in the mid 1970's. The Israeli's success led to Secretary of the Navy Lehman's 
impression that a useful RPV capability could be quickly acquired. The threat faced by the 
Navy is not the same as that in the NATO batUefields. The Navy and Marine Corps 
acquired several PIONEER systems, before Congress prohibited further Service RPV 
procurements.^^ Congress and DoD, favorably impressed by the Navy's progress, have 
given the Navy responsibility for running the DoD RPV Joint Program 
PIONEER, however, is not in the competition for the fumrc joint-Service shon-range 
RPV.^i It is expected to be superseded by other designs. 

The AQUILA anti-jam communications systems (MICNS) was developed by the 
same contractor (Hauls) which had made the earUcr ICNS used in PRAEIRE. About $2 
million was spent by DARPA on the integrated communications and navigation system 
(ICNS) and about $100 million by the Anny on MICNS. Trade-offs have had to be made 
between space and weight on RPVs, and antijam capabiUty which depends on the 

Difficulty has persisted with IR technology for the mini-RPVs. ARPA had 
problems with the early CALERE and AQUILA at the time of cancellation did not have a 
satisfactory package.^^ 

DARPA's reentry into RPVs, the AMBER program, was oriented to larger RPVs 
with long endurance, low observables and sophisticated sensor technology. AMBER has 
been transferred to the Services, The Boeing-developed CONDOR, recentiy supported by 
DARPA, has aroused considerable interest in the Army and Navy. 

The DARPA outlay for mini-RPVs, between 1972 and 1977, was nearly $15 
million.5* The Army's outlays for AQUILA were, at the time of cancellation, about $800 

*9 "Pentagon Considers Buying Additional Pioneer RPVs." by John D. Morocco. Aviation 
July 31, 1989, p. 81. 

50 Discussion with J. Kreis, 8/89. 

5 1 Aviation Week, ibid. 

52 GAO Report, ibid. 

53 The last IR payload contractor for AQUILA was Foid, which had buUt the FLIRs for CALERE. 

54 Hearings, ibid. Testimony of K. Kicsa. p. 3974. 


miUion doUars.55 By im(i-1989 the Navy and Marine Corps had procured nine PIONEER 
systems, at a cost of about $63 miUion.56 The DoD UAV Joint Program Office is expected 
to have a budget of some $50 miUion/year when it can produce a coordinated plan to satisfy 
Congress. However, the formation of this office and its primary concern with RPV 
acquisition has led to reduction of the DARPA RPV effort^*' 

55 Hearing before the Commiuee on Anncd Services. Department of Defense Authonzation for 
Appropriations. By 1987, 94th Congress, 2nd Session, RDT&E, Tide II, Tes&mony of Gen. Wagner, 
p. 807. 

56 "Pentagon Considers Buying Additional PIONEER RPVs," by John D. Morocco. Aviation Week. 
July 31, 1989, p. 81. 

57 "DARPA May Use Boeing Drone for Prototype," Aviation Week, Nov. 28, 1988, p. 86. 




















USD 1-5 






t • • • • • pRAEIRE 






I* • 

• • • FLIR 



1988 I 






J . 




73 WAR 


82 WAR